Intelligent life buoy unmanned aerial vehicle

Abstract

An intelligent life buoy unmanned aerial vehicle comprises an unmanned aerial vehicle body and a life buoy which are connected through a rope and a claw buckle, the unmanned aerial vehicle body comprises an intelligent chip, a waterproof rotor wing, an infrared camera, a microphone, a GPS positioning module, a wireless communication module and a remote controller, and the intelligent life buoy unmanned aerial vehicle is characterized by adopting the following working steps that 1, the intelligent life buoy unmanned aerial vehicle body recognizes the position of a drowning person; dragging the life buoy to be close to the drowning person in a vertical posture by using a rope; 2, the intelligent life buoy unmanned aerial vehicle recognizes the face position of the drowning person, a rope is dragged in the opposite direction of the face position of the drowning person, and the life buoy is controlled to turn downwards around the head of the drowning person to sleeve the drowning person; 3, the intelligent life buoy unmanned aerial vehicle drags the rope and the life buoy to drag the sleeved drowning person back to the shoreside in the upward posture. The life buoy can be sleeved on a drowning person, and by means of buoyancy of the life buoy and lift force of the unmanned aerial vehicle, the drowning person can be provided with a stronger escape ability by means of the buoyancy of the life buoy and the lift force of the unmanned aerial vehicle.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more particularly to lifebuoys, drones, and artificial intelligence technology, specifically to intelligent lifebuoy drones. Background Technology

[0002] High-risk groups for drowning include children who lack knowledge about water safety, and those who swim in dangerous waters or enter the water without permission. The main causes of drowning deaths include swimming alone, improper rescue, and insufficient awareness of water risks. Once a drowning incident occurs, timely rescue becomes the most important task, commonly involving lifeguards, deploying lifebuoys, dispatching rescue boats, and rescue aircraft.

[0003] In recent years, the rapid development of drone technology has led to significant progress in multiple fields, particularly in flight control, sensor technology, communication capabilities, intelligence levels, and the expansion of application scenarios. Currently, advanced flight control systems and sensor technologies enable drones to hover and fly more stably, maintaining a stable attitude even in strong winds. Flight speeds have also increased considerably, allowing for rapid arrival at designated locations and flexible adjustments to attitude and direction during flight. Some drones also possess the ability to quickly switch flight modes to adapt to different mission requirements. Modern drones can also be equipped with various sensors such as high-definition cameras, thermal imagers, gas sensors, and lidar, enabling rapid and accurate monitoring and assessment of different types of disasters. For example, in the event of a toxic gas leak, drones can carry gas sensors into the danger zone to monitor the spread of toxic substances in real time. In terms of communication, the integration with 5G communication technology has greatly enhanced the communication capabilities and data transmission speeds of drones. For instance, the "Wing Loong" drone, equipped with 5G 700MHz emergency communication technology, can provide stable communication coverage in extreme conditions, supporting voice communication and SMS messaging.

[0004] With the development of artificial intelligence and machine learning technologies, drones have gained stronger capabilities in autonomous decision-making, navigation, and obstacle avoidance. For example, through deep learning algorithms, drones can autonomously identify complex environments and objects, and perform precise path planning and obstacle avoidance. The application areas of drones are also constantly expanding, from traditional military, agricultural, and monitoring fields to emerging fields such as logistics, film and television production, aerial tourism, and emergency rescue. In the field of emergency rescue, especially in water rescue, drone technology has become an important innovative tool. For example, invention CN120182858A discloses a drone-based method for detecting maritime search and rescue targets based on an improved YOLOV8 algorithm. Currently, drone lifebuoy technology has the following significant characteristics and advantages: improved delivery technology for rescue equipment; existing drones can carry lifebuoys, inflatable rafts, and other equipment, which can be released via robotic arms or airdrop mechanisms. Some systems integrate GPS positioning and wireless communication (4G / 5G) to achieve remote monitoring and collaborative rescue; and they possess advanced sensors and positioning systems: modern drones are equipped with high-resolution cameras, infrared imaging equipment, lidar, and other advanced sensors, enabling them to accurately locate drowning victims in complex environments. These devices can be used not only during the day, but also at night or in low visibility conditions; by fusing multiple sensors and combining technologies such as infrared thermal imaging and millimeter-wave radar, they can improve target detection capabilities in adverse weather conditions (fog, rain).

[0005] Despite significant progress made in the field of water rescue in recent years, some problems and shortcomings still need to be addressed: 1. Low rescue efficiency. The golden time for drowning rescue is within 4-6 minutes, during which brain hypoxia is still in a reversible stage. If this time is exceeded, brain cells will die more quickly due to hypoxia, significantly reducing the survival rate and the rate of neurological function recovery. For every minute of delay in rescue, the probability of survival decreases by about 10%, therefore, every second counts in rescue efforts. Although lifebuoys are an effective rescue tool, they are often unavailable at rescue sites. It is also impossible to deploy them to distant bodies of water quickly enough by manpower alone, and it is also impossible to ensure they are deployed vertically. Lifebuoys can only be properly fitted when deployed vertically, with the drowning person's hands and head passing through the holes. If the drowning person only grabs the lifebuoy without putting it on, they are easily dislodged in the face of strong currents, large waves, cold weather, or exhaustion. Furthermore, drowning victims often fail to use lifebuoys correctly in the chaos of an emergency, making it difficult to properly slip them on. Firstly, standing a lifebuoy upright requires considerable effort, which drowning victims often lack. Secondly, an upright lifebuoy provides minimal buoyancy; without additional buoyancy or lift, the drowning victim will likely choke on water when trying to put the lifebuoy on. Traditionally, rescue operations in more distant waters involve lifeboats, but these are slow, and the vessels themselves pose safety threats to the drowning victim, such as the potential for injury from the propeller.

[0006] 2. Complex Rescue Environments. Many rescue environments are extremely complex. For example, in situations far from shore with large waves, rescuers are often helpless. Drowning victims who receive lifebuoys are often exhausted and unable to return, let alone when encountering large waves, going against the current, or being very far from shore. In such cases, long-distance and accurate deployment of lifebuoys is almost impossible. In addition, rip currents, undercurrents, eddies, seaweed, and underwater creatures can all alter the drowning victim's environment, making accurate deployment of lifebuoys impossible or even causing them to be swept away by the current. This also poses a significant threat to the personal safety of lifeguards, and makes the operation of rescue boats and rescue aircraft extremely difficult.

[0007] 3. Drone rescue technology is immature. In complex environments, the accuracy of drones deploying lifebuoys is insufficient, delaying rescue time and efficiency; communication and signal issues exist, as drone communication signals are easily interrupted in complex terrain or marine environments, leading to difficulties in command and control. Signal interference may also affect drone flight safety; safety concerns arise, as drones may cause collisions and injuries to those being rescued during the rescue process. Furthermore, existing technologies mostly focus on single functions (such as detection or delivery) and have not yet formed a complete "perception-decision-rescue" closed-loop system. Summary of the Invention

[0008] The intelligent lifebuoy drone proposed in this invention combines lifebuoys, drones, and artificial intelligence technology to provide an intelligent drowning rescue platform, further improving the efficiency of drowning rescue.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A smart lifebuoy drone includes a drone and a lifebuoy connected by a rope and a claw buckle. The drone includes a smart chip, waterproof rotors, an infrared camera, a microphone and speaker, a GPS positioning module, a wireless communication module, and a remote controller. The drone operates using the following steps: Step 1: The intelligent lifebuoy drone identifies the location of the drowning person and uses a rope to drag the lifebuoy vertically towards the drowning person; Step 2: The intelligent lifebuoy drone identifies the drowning person's face position, drags the rope in the opposite direction of the drowning person's face position, and controls the lifebuoy to flip down around the drowning person's head and cover the drowning person; Step 3: The intelligent lifebuoy drone drags the rope and lifebuoy to pull the trapped drowning person back to shore in a supine position.

[0010] Step 1 includes the following sub-steps: Sub-step 1-1: Rescuers send rescue commands to the smart lifebuoy drone via remote control; Sub-steps 1-2: The intelligent lifebuoy drone identifies the location of the drowning victim and flies towards them; Sub-steps 1-3: When close to the drowning person, the intelligent lifebuoy drone releases the claw buckle and lowers the lifebuoy, using only the rope to drag the upper part of the lifebuoy to ensure that the lifebuoy is in a vertical position; Sub-steps 1-4: The intelligent lifebuoy drone identifies the drowning victim's face and controls the rope to drag the lifebuoy in a vertical position towards the drowning victim's face and head.

[0011] Step 1-1 includes the following sub-steps: Sub-step 1-1-1: Rescuers activate the waterproof rotors of the smart lifebuoy drone via remote control, and the smart lifebuoy drone takes off. Rescuers then activate the infrared camera via remote control to capture infrared images of the drowning victim at the scene. Sub-step 1-1-2: Rescuers use a remote control to mark the location of the drowning victim with a red frame in the infrared image of the drowning victim and send rescue instructions; Steps 1-2 include the following sub-steps: Sub-step 1-2-1: The intelligent lifebuoy drone uses infrared image guidance to track the drowning person. The intelligent chip controls the flight speed, altitude and direction of the intelligent lifebuoy drone by adjusting the waterproof rotor, aiming at the position of the drowning person marked by the red box in the infrared on-site image. Upon receiving the rescue order, the intelligent lifebuoy drone adjusts its flight speed, altitude, and direction, while simultaneously activating its infrared camera, microphone, speaker, and GPS positioning module. Sub-step 1-2-2: The intelligent lifebuoy drone continuously captures infrared images of the drowning victim and adjusts the angle of the infrared camera to update the position of the drowning victim marked by the red box in the infrared image, ensuring that the position of the drowning victim marked by the red box is located in the center of the infrared image. The smart chip acquires images from the infrared camera, sound from the microphone and speaker, and real-time location from the GPS positioning module, and sends them to the remote control via the wireless communication module. Sub-steps 1-2-3: Rescuers can use the smart lifebuoy drone as a rescue drone and start another smart lifebuoy drone as an observation drone via remote control; the observation drone continuously captures infrared images of the rescue drone and the drowning victim, updates the position of the drowning victim marked in red in the infrared images, and the observation drone can correct the position, direction and speed of the rescue drone through the wireless communication module; If there are multiple drowning victims, the operator can use a remote control to directly operate multiple smart lifebuoy drones to fly to each drowning victim and carry out rescue operations on multiple victims simultaneously.

[0012] In step 1-2-1, the operation of infrared image guidance includes the following steps: Sub-step 1-2-1-1: The smart chip activates the infrared camera to capture infrared images of the drowning victim and activates the microphone and speaker to collect sounds from the drowning victim. Sub-step 1-2-1-2: The smart chip runs a preprocessing algorithm to filter, convert the format, identify the target, and extract features from the infrared images of the drowning victim captured by the infrared camera. The intelligent chip reduces image noise through Gaussian filtering and preprocesses the real-time acquired infrared images of drowning victims using the following formula: ; in, It is a Gaussian kernel, and k is the kernel size. The image shows an infrared image of a drowning victim, where x and y are the pixels in the horizontal and vertical directions of the infrared image, respectively. Sub-steps 1-2-1-3: The smart chip runs a preprocessing algorithm to filter, convert the format of, identify targets and extract features from the on-site sounds of the drowning victim collected by the microphone and speaker; Sub-step 1-2-1-4: The intelligent chip processes the infrared on-site image frames of the drowning victim preprocessed in sub-step 1-2-1-2. As input, the resolution is W H C, representing width W, height H, and number of channels C, represents the infrared on-site image frames of the drowning victim as a time series. ,in It is the t-th frame; The intelligent chip adjusts the size of the infrared image frames of the drowning victim to fit the model's input, using the following formula: ; in These are the model input dimensions; Sub-steps 1-2-1-5 involve the intelligent chip normalizing the pixel values ​​of the infrared image frames of the drowning victim, using the following formula: ; in and These are the mean and standard deviation; In sub-step 1-2-1-6, the intelligent chip uses the preprocessed drowning victim audio from sub-step 1-2-1-3 as input. The input drowning victim audio is... Representing the sounds of a drowning victim as a time series ,in It is the sound at time t; In sub-step 1-2-1-7, the intelligent chip extracts features from the infrared image frames of the drowning victim in sub-step 1-2-1-5 and the sounds of the drowning victim in sub-step 1-2-1-6. A convolutional neural network is used to extract the features of the infrared image frames and the sounds of the drowning victim. Here, the CNN is set to have L layers, and the output formula of the l-th layer is: ; in It is the output of the l-th layer. It is the output of the ll-th layer. Convolutional kernel weights It's a bias. It is an activation function. Indicates the convolution operation; Sub-steps 1-2-1-8 output the feature map F of the infrared on-site image frame of the drowning victim, with a size of [missing information]. × × ,in , , These are the feature dimensions: width W, height H, and number of channels C. Sub-steps 1-2-1-9: The smart chip performs drowning target detection based on the sounds at the drowning site, calculates whether the feature map of the drowning site sounds has the sound feature of calling for help, calculates the matching probability of the sound feature of calling for help, and sets it as the sound confidence level C1. Sub-step 1-2-1-10: The intelligent chip performs drowning target detection based on the feature map F of the drowning victim's infrared on-site image frame, dividing the drowning victim's infrared on-site image frame into... A grid is generated, with each grid predicting B bounding boxes, and each bounding box containing center coordinates. Given the width w and height h, calculate whether the feature map of the infrared on-site image frame of the drowning victim has the image features of the drowning victim, calculate the probability of the drowning victim's existence, and set it as the image confidence level C2; Sub-step 1-2-1-11: The intelligent chip calculates the category probability P(c|object), representing the probability that the target belongs to category C2. ,in It is the image confidence C2 category score; In sub-step 1-2-1-12, based on the sound confidence C1 from sub-step 1-2-1-9 and the image confidence C2 from sub-step 1-2-1-10, the intelligent chip outputs the drowning victim target detection formula as follows: ; In sub-step 1-2-1-13, the intelligent chip further calculates the total confidence C and marks the bounding box of the drowning victim in the infrared image frame of the drowning victim, which represents the probability that the bounding box contains the drowning victim target and the prediction is correct: ; in It is the intersection-union ratio of the predicted bounding box and the ground truth bounding box, which can be obtained by statistical analysis based on historical data; Sub-step 1-2-1-14: The intelligent chip performs non-maximum suppression to remove overlapping bounding boxes from the infrared image frames of the drowning victim, retaining the boxes with the highest confidence, and calculating the intersection-union ratio (IUU). The formula is: ; ,if If true, then remove The bounding box retains the confidence level. The threshold value is 0.5. Sub-step 1-2-1-15: The intelligent chip marks the drowning victim target detected in the infrared on-site image frame of the drowning victim with a red bounding box. The coordinates of the bounding box are: Draw a rectangle; Sub-step 1-2-1-17: The smart chip adds a type label and confidence level above the bounding box of the drowning victim in the infrared on-site image frame of the drowning victim; In sub-step 1-2-1-18, the intelligent chip estimates the distance from the intelligent lifebuoy drone to the drowning victim based on the shape and size of the drowning victim marked in red in the infrared live image and the human body size in historical infrared live image data. Then, based on the drowning victim target from sub-step 1-2-1-15, it determines whether the distance estimation meets the error condition. If it does, the coordinates of the drowning victim are located. ,in It is the estimated three-dimensional coordinates of the drowning victim in longitude, latitude, and altitude; Sub-step 1-2-1-19: The smart chip locates the drowning victim's coordinates based on step 1-2-1-18. Step 1-2-1-15: Real-time image acquisition Output the bounding box of the drowning person To accurately locate a drowning victim, the monocular ranging formula is as follows: ; in, It is the actual height of the drowning victim. is the actual altitude of the intelligent lifebuoy drone, and f is the focal length of the camera; Sub-step 1-2-1-20: The intelligent chip calculates the relative coordinates of the intelligent lifebuoy drone using the following formula: ; in, , These are the optical center coordinates in longitude and latitude. and It refers to the focal length in terms of longitude and latitude; Sub-step 1-2-1-21: The intelligent chip obtains the coordinates of the intelligent lifebuoy drone's location. The total Euclidean distance between the intelligent lifebuoy drone and the drowning victim is calculated using the following formula: ; Sub-step 1-2-1-22: The intelligent chip performs a global path planning algorithm to select the path from the intelligent lifebuoy drone to the drowning person, and calculates the actual cost of the intelligent lifebuoy drone flying from the starting position to the drowning person using the following formula: ; in, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-23, based on the Euclidean total distance between the intelligent lifebuoy drone and the drowning person in sub-step 1-2-1-21. The actual cost in sub-step 1-2-1-22 The intelligent chip further obtained the cost function for the intelligent lifebuoy drone to fly to the drowning victim's location as follows: ; Sub-step 1-2-1-24: The elected intelligent lifebuoy drone samples the nodes of each path that can reach the drowning person, gradually constructing a tree structure, and the optimized path formula is: ; in, It is the smoothing coefficient. Obstacle avoidance loss, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-25: The smart chip performs distributed information sharing. Each smart lifebuoy drone broadcasts its own status, including the distance to the drowning person, the percentage of battery power, and the real-time location collected by the GPS positioning module, and uses time division multiple access to prevent channel congestion. Sub-step 1-2-1-26: The smart chip elects the nearest smart lifebuoy drone, that is, all smart lifebuoy drones elect the smart lifebuoy drone closest to the drowning person to carry out the rescue. All the smart lifebuoy drones are constantly changing positions. The primary strategy of dynamic election is to select the smart lifebuoy drone that is closest to the drowning person. The secondary strategy is to select the smart lifebuoy drone with the highest battery percentage if the distance difference between multiple smart lifebuoy drones is less than a certain distance. The selected smart lifebuoy drone broadcasts confirmation, while the other smart lifebuoy drones switch to observation mode. Sub-step 1-2-1-27: The selected intelligent lifebuoy drone needs to construct a local map, and simultaneously construct a local map and accurately locate the drowning person on the map based on the GPS positioning module and infrared camera. Sub-step 1-2-1-28: The intelligent chip performs motion control on the flight of the intelligent lifebuoy drone, and the PID output of the waterproof rotor... The control command formula is: ; in For time variables, Represents proportional, integral, and derivative gain, and output error. Locating the coordinates of a drowning victim Coordinates of the location of the intelligent lifebuoy drone .

[0013] Sub-step 1-2-1-29: The intelligent chip predicts and controls the flight path of the intelligent lifebuoy drone towards the drowning victim, and the formula for selecting the optimal path is: ; in, It is a predicted future state. It controls the input sequence; In sub-step 1-2-1-30, if the distance to the drowning victim does not meet the preset threshold, return to sub-step 1-2-1-18 to re-obtain the drowning victim's coordinates. Step 1-2-1-21: Reacquire the coordinates of the intelligent lifebuoy drone's current location. This is to prevent the drowning person's position from changing during the rescue process; Sub-step 1-2-1-31: The smart chip performs anomaly monitoring. If the drowning victim is lost for more than a certain period of time, a re-election is triggered, and the process returns to sub-step 1-2-1-1. Sub-step 1-2-1-32: If the smart chip detects that the smart lifebuoy drone's battery percentage is below a certain value, it will return to home and request a replacement smart lifebuoy drone from the remote controller.

[0014] Steps 1-3 include the following sub-steps: Sub-step 1-3-1: The intelligent lifebuoy drone estimates the distance between itself and the drowning person based on the shape and size of the drowning person marked in red in the infrared live image and the human body size in historical infrared live image data. Preferably, the remaining flight time between the two is used to estimate the distance between them. The intelligent lifebuoy drone estimates the remaining flight time based on the rate of change of the shape and size of the drowning person marked in red in the infrared live image and the time already flown. The intelligent lifebuoy drone uses infrared image guidance to track the lifebuoy as it approaches the drowning person; Sub-step 1-3-2: If the intelligent chip estimates that the remaining flight time is short, the waterproof rotor is adjusted to gradually reduce the flight speed and altitude; preferably, when the intelligent chip estimates that the remaining flight time is 1-2 minutes, it is closer to the drowning person. Gradually reducing the flight speed is necessary to ensure that when the intelligent lifebuoy drone reaches the drowning person, its flight speed can be reduced to 0, hovering above the drowning person so that the drowning person can grab the lifebuoy; Gradually reducing the flight altitude requires ensuring that the lower part of the lifebuoy can touch the water surface when it is lowered to buffer the descent speed of the lifebuoy and avoid excessive descent speed of the lifebuoy drone, which would excessively interfere with the normal flight of the lifebuoy drone by pulling it with the rope. Sub-step 1-3-3: When the intelligent chip estimates that the flight altitude reduction meets the requirements, it controls the release of the claw buckle to lower the lifebuoy. The drone and the lifebuoy separate, the lifebuoy falls and its lower part touches the water surface, and the upper part of the lifebuoy is kept connected to the drone by a rope.

[0015] Steps 1-4 include the following sub-steps: Sub-step 1-4-1: The smart chip captures infrared images of the drowning victim using an infrared camera and identifies the victim's face. Sub-step 1-4-2: The smart chip determines whether the current flight direction of the smart lifebuoy drone is aligned with the face and head of the drowning person; if so, it continues to fly in the direction of the face and head of the drowning person; if not, the smart chip adjusts the waterproof rotor to change the current direction to fly in the direction of the face and head of the drowning person. Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the face and head of the drowning person, and ensures that the face and head of the drowning person are located in the center of the infrared on-site image.

[0016] Step 2 includes the following sub-steps: Sub-step 2-1: The intelligent lifebuoy drone hovers above the drowning person, allowing the drowning person to grab the lifebuoy, and identifies whether the drowning person has grabbed the lower part of the nearest lifebuoy; Sub-step 2-2: If the drowning person manages to grab the lower part of the nearest lifebuoy, the intelligent lifebuoy drone identifies the drowning person's face position and flies behind the drowning person. It controls the rope to drag the upper part of the lifebuoy behind the drowning person and flip it to a horizontal position, so that the lifebuoy can be put around the drowning person from head to toe. Sub-steps 2-3: The intelligent lifebuoy drone identifies whether the lifebuoy is horizontal and whether the drowning person's face and hands are on the lifebuoy, confirming that the lifebuoy has successfully encased the drowning person.

[0017] Step 2-1 includes the following sub-steps: Sub-step 2-1-1: The smart chip captures an infrared image of the drowning victim using an infrared camera and identifies whether the victim can touch the lifebuoy. If so, it maintains its current position and hovers above the victim. If not, the smart chip adjusts the waterproof rotor to control the lifebuoy to approach the victim's face at an extremely low flight speed. Furthermore, if there is wind interference at the scene causing the rope to drag the lifebuoy, resulting in the lifebuoy rotating or deviating from the victim's face, the smart chip adjusts the waterproof rotor to change its current speed and direction, flying towards the victim's face and head. The extremely low flight speed is no more than 1 meter per second.

[0018] Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy as it approaches the drowning person's face, ensuring that the drowning person's face is located in the center of the infrared image. Sub-step 2-1-2: The smart chip obtains the current GPS coordinates of the drowning person through the GPS positioning module and reports the current GPS coordinates of the drowning person to the remote control through the wireless communication module; the smart chip picks up the sound at the scene through the microphone and speaker, or plays voice messages to the scene to comfort the drowning person and guide the drowning person to use the lifebuoy correctly. Sub-step 2-1-3: Since the lifebuoy is vertically positioned close to the water surface, the drowning person can only grab the bottom of the lifebuoy and allow their head and hands to naturally enter the lifebuoy's holes. The smart chip uses an infrared camera to capture an infrared image of the drowning person and identifies whether the drowning person has already grabbed the bottom of the lifebuoy or whether the drowning person's head and hands have naturally entered the lifebuoy's holes. If so, proceed to sub-step 2-2; otherwise, proceed to sub-step 2-1-1. Furthermore, if there is wind interference at the drowning person's location causing the rope to drag the lifebuoy, resulting in the lifebuoy rotating or deviating from the drowning person's face, once the drowning person has grabbed the bottom of the lifebuoy, the rotation or deviation of the lifebuoy will disappear on its own, and the smart chip will not need to adjust its flight speed and direction. Step 2-2 includes the following sub-steps: Sub-step 2-2-1: The smart chip captures infrared images of the drowning victim using an infrared camera, identifies the victim's face position, adjusts the waterproof rotor to fly at an extremely low speed behind the drowning victim, and controls the rope to drag the upper part of the lifebuoy to flip behind the drowning victim. The intelligent chip adjusts the waterproof rotors, controlling the intelligent lifebuoy drone to perform flipping flight with the lower part of the lifebuoy as the center and the sum of the lifebuoy's diameter and the rope's length as the radius; Sub-step 2-2-2: The smart chip captures an infrared image of the drowning victim using an infrared camera, identifies the position of the lifebuoy and the drowning victim's head, and determines whether the lifebuoy has flipped over the drowning victim's head. If so, the lifebuoy has successfully wrapped around the drowning victim, and the process proceeds to sub-step 2-3. If not, the lifebuoy has not successfully wrapped around the drowning victim, and the process proceeds to sub-step 2-2-1 to continue the flipping operation.

[0019] Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy's rotation around the drowning person, ensuring that the drowning person's head is located in the center of the infrared image. In steps 2-3, the smart chip captures an infrared image of the drowning victim using an infrared camera, and identifies the lifebuoy and the drowning victim's head and hands. It determines whether the drowning victim's head and hands are on the lifebuoy, whether the lifebuoy image is nearly a perfect circle, and whether the connection between the lifebuoy and the rope is on the drowning victim's back. If so, the lifebuoy is horizontal, successfully securing the drowning victim, and the rope pulls the lifebuoy from the drowning victim's back, proceeding to step 3. If not, the drowning victim has not been successfully secured, and the process proceeds to sub-step 2-2-1 to continue the flipping operation.

[0020] Step 3 includes the following sub-steps: Sub-step 3-1: The intelligent lifebuoy drone identifies and returns to the shore location; Sub-step 3-2: The intelligent lifebuoy drone plans the optimal flight route; Sub-step 3-3: The intelligent lifebuoy drone follows the optimal flight path, dragging the rope and lifebuoy to pull the trapped drowning person back to shore, keeping the lifebuoy horizontal throughout the process.

[0021] Step 3-1 includes the following steps: Sub-step 3-1-1: Rescuers send the GPS coordinates of the return-to-shore position to the intelligent lifebuoy drone via remote control; Sub-step 3-1-2: The intelligent lifebuoy drone acquires and identifies the GPS coordinates of the return shore location; Step 3-2 includes the following steps: Sub-step 3-2-1: The intelligent lifebuoy drone obtains the GPS coordinates of its current location through the GPS positioning module; Sub-step 3-2-2: The intelligent lifebuoy drone plans the optimal flight route based on the GPS coordinates of the return shore location, the current location, and the water imagery and obstacle conditions between the two, so that the drowning victim can avoid dangers and obstacles on the return journey. The intelligent lifebuoy drone uses infrared image guidance for path planning and monitors the drowning victim's condition throughout the process. Step 3-3 includes the following steps: Sub-step 3-3-1: The intelligent lifebuoy drone flies along the optimal flight path, ensuring that the rope pulls the lifebuoy from the drowning person's back, and the drowning person returns to the shore with their back facing the shore in a backstroke position; the intelligent lifebuoy drone monitors the drowning person's condition throughout the process to ensure minimal water inhalation and maximum safety, with rescuers and the intelligent lifebuoy drone working together to ensure the safety of the drowning person's return route.

[0022] Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the drowning person's status and ensures that the drowning person's head is located in the center of the infrared on-site image. Sub-step 3-3-2: The rescue is completed. The intelligent lifebuoy drone returns to the standby point, awaits new instructions, and uses the claw buckle to re-secure the lifebuoy.

[0023] Compared with the prior art, the present invention has the following technical effects: 1. This invention uses a drone to control a lifebuoy that is directly placed on the drowning victim, eliminating concerns about losing the lifebuoy and improving the efficiency of water rescue. The invention allows the lifebuoy to fly vertically towards the drowning victim, with the drone completing the placement process. When the drowning victim's hand touches the bottom of the lifebuoy, it flips to a horizontal position and secures the victim, perfectly solving the problems of panicked victims not knowing how to use the lifebuoy correctly, and victims lacking the strength to hold it upright. The intelligent lifebuoy drone can intelligently patrol, respond quickly, and accurately locate drowning victims. Furthermore, the drone itself is a lifebuoy, allowing it to quickly land next to the drowning victim without requiring a lifeguard to enter the water. Moreover, if there are multiple drowning victims, the operator can remotely control multiple intelligent lifebuoy drones to fly to each victim simultaneously, enabling simultaneous rescue of multiple victims. 2. This invention uses a drone to tow a lifebuoy to rescue a drowning victim back to shore, without expending the victim's energy, resulting in rapid rescue. This invention is suitable for very complex rescue environments, such as encountering large waves or going against the current, far from the shore, in cold, snowy, or weedy conditions. Drowning victims who have received the lifebuoy are often exhausted, unconscious, and unable to return, leaving rescuers helpless. The intelligent lifebuoy drone patrols above the water and intelligently identifies the drowning victim, achieving a rescue speed far exceeding traditional rescue techniques. The intelligent lifebuoy drone combines information age technology with its own capabilities, enabling joint optimization even under information age limitations. It integrates wireless communication technology, artificial intelligence algorithms, and an autonomous flight control system, further enhancing rescue speed. 3. This invention features a lifebuoy that is always attached to a rope, preventing loss and allowing for repeated deployments with precise rescue. Unlike traditional lifebuoys or drone-borne lifebuoys, which are easily lost and difficult to retrieve if misdelivered or if the drowning person loses their grip, this invention uses a drone that remains airborne and controlled by a remote, unaffected by water currents. Facial recognition and the rope-attached lifebuoy ensure accurate and controllable deployment. Even if the lifebuoy is misdelivered initially, the drone can quickly adjust its position by dragging the rope and deploy it repeatedly. During the patrol, a smart chip drives a waterproof camera to capture photos or videos of the water surface and uses a microphone to capture sound above the water. The combination of these elements helps determine the drowning person's condition and location. The intelligent lifebuoy drone incorporates multiple safety features, such as ultra-lightweight materials like foam or air filling, to ensure no harm is caused to the drowning person during the rescue.

[0024] 4. This invention utilizes the buoyancy of a lifebuoy and the lift of a drone to provide drowning victims with a significantly stronger escape capability, and it can operate with multiple drones intelligently. Since the lift of a drone exceeds the weight of an adult, when a drowning victim is heavy and encounters rapids, currents, cold weather, floods, or entanglement in aquatic plants, this invention can provide a far greater escape capability than a lifebuoy. Furthermore, multiple intelligent lifebuoy drones can communicate collaboratively, exchange and share information, coordinate patrols and rescues, and automatically select the closest intelligent lifebuoy drone to the drowning victim for rescue, further improving rescue efficiency. By integrating autonomous navigation, high-precision delivery, and rescue design, the intelligent lifebuoy drone solves the efficiency and reliability bottlenecks of traditional rescue technologies, providing a revolutionary solution for water safety. Attached Figure Description

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a system structure diagram of the intelligent lifebuoy drone in this invention.

[0026] Figure 2This is a flowchart illustrating the working method of the intelligent lifebuoy drone in this invention.

[0027] Figure 3 This is a schematic diagram of step 1 of the intelligent lifebuoy drone in this invention.

[0028] Figure 4 This is a schematic diagram of step 2 of the intelligent lifebuoy drone in this invention.

[0029] Figure 5 This is a schematic diagram of step 3 of the intelligent lifebuoy drone in this invention. Detailed Implementation

[0030] like Figure 1 The diagram shows a system structure of an intelligent lifebuoy drone, comprising a drone 100 and a lifebuoy 200 connected by a rope 300 and a claw buckle 400. The drone 100 includes an intelligent chip 101, a waterproof rotor 102, an infrared camera 103, a microphone and speaker 104, a GPS positioning module 105, a wireless communication module 106, and a remote controller 107. The intelligent chip 101 is connected to the claw buckle 400. The drone 100 is the main structure of the intelligent lifebuoy drone, providing lift for the present invention. Preferably, it is a quadcopter or hexcopter high-power drone, capable of carrying the national standard 4.3KG professional lifebuoy mentioned in the drone 100. Both the quadcopter and hexcopter have protective covers and strong waterproof capabilities. Preferably, it features optical flow positioning hovering, brushless power, a built-in rechargeable battery, a 1-hour flight time, 2.4G signal transmission, and a load capacity exceeding 75KG.

[0031] The intelligent chip 101, installed inside the drone 100, is one or two in number and connected to the infrared camera 103 and the waterproof rotor 102. It can collect photos or videos captured by the infrared camera 103, read operation commands sent by the remote controller 107, or issue its own operation commands to drive the waterproof rotor 102, thereby controlling the flight speed and direction of the intelligent lifebuoy drone. The intelligent chip 101 provides the core computing power for this invention, running artificial intelligence algorithms. Preferably, it is an intelligent accelerator card equipped with an ultra-thin AE7100 chip, enabling fully programmable design, 10W power consumption, LPDDR4 / 4x 128bit memory, 8GB / 16GB capacity, and compatibility with CUDA, ONNX, Runtime API, etc. It supports models including Llama2, Stable Diffusion, YOLOv5, ResNet, etc., with a computing performance of INT8. With a computing power of 25.6 TOPS and a structural size of 80mm in length and 22mm in width, its compact size provides high flexibility to handle various AI applications. It boasts a computing power of up to 25.6 TOPS and a memory bandwidth of 60GBs, ensuring efficient and stable processing and data transmission.

[0032] The waterproof rotor 102 is the power system of the intelligent lifebuoy drone. It is installed around the drone 100 and can provide lift to the intelligent lifebuoy drone through high-speed rotation and control the direction of travel of the intelligent lifebuoy drone. There are four or more of them. Preferably, the waterproof rotor 102 can also work normally when it falls on the water surface. The infrared camera 103, mounted on the outside of the drone 100, is used to observe the water surface and capture photos or videos. One or two cameras are used; preferably, the infrared camera 103 can function normally even when submerged in water. The infrared camera 103 provides depth images and target detection for both day and night, and can function normally even when submerged. Preferably, it is a waterproof infrared high-definition camera with a 1 / 2.7" Progressive Scan CMOS sensor, a minimum illumination of 0.002 Lux @ (F1.2, AGC ON) for color, 0 Lux infrared LED on, infrared distance ≥ 50 meters, shutter speed 1 / 3 s ~ 1 / 100,000 s, wide dynamic range 120 dB, day / night switching mode, ICR infrared filter type, focal length & field of view 2.7 ~ 12 mm, horizontal field of view: 110° ~ 35°, vertical field of view: 58° ~ 20°, diagonal field of view: 132° ~ 41°, maximum aperture F1.2, and maximum image size 1920. ×1080. Furthermore, it features intelligent surveillance, including area intrusion detection, boundary crossing detection, area entry detection, area departure detection, item abandonment detection, item retrieval detection, loitering detection, parking detection, crowd gathering detection, and rapid movement detection.

[0033] The microphone and speaker 104 are mounted on the outside of the drone 100 and are used to pick up on-site sound or play voice messages on-site. They are capable of recognizing distress calls and can be used in one or two configurations. Preferably, the microphone and speaker 104 can function normally even when submerged in water. The microphone and speaker 104 provide sound input and output functions for this invention. Preferably, the microphone is an externally polarized condenser microphone with a built-in 22mm gold-plated large diaphragm pickup head and an equivalent noise level of 15dB, ensuring high sound reproduction and detail. Preferably, the speaker is 4.5W, with a sound source response time of less than 250ms, a maximum volume of 95dB, and is adjustable via both software and hardware. The operating temperature is -40℃ to 60℃.

[0034] The GPS positioning module 105, installed inside the drone 100, consists of one module. It accurately locates the current position of the intelligent lifebuoy drone and sends the information to the intelligent chip 101 and the remote controller 107. The GPS positioning module 105 provides accurate positioning for this invention. Preferably, it is a high-precision, low-power quad-mode satellite positioning module, i.e., single-mode, dual-mode, and multi-mode operation of BeiDou + GPS + Galileo + GLONASS, which can be switched between modes via commands. The operating voltage is 3.0V~3.5V, supports SBAS, QZSS, and A-GNSS assisted positioning, has an acquisition sensitivity of -147dB, a tracking sensitivity of -163dB, a positioning accuracy of less than 3 meters, supports power supply to an active antenna, and has an operating temperature of -40℃~85℃.

[0035] The remote controller 107, operated by the operator, can communicate with the intelligent lifebuoy drone, send operation commands to the intelligent lifebuoy drone, and receive photos or videos sent by the intelligent lifebuoy drone. The remote controller 107 provides remote control for this invention. Preferably, the remote control distance is greater than 300 meters, with real-time screen transmission, six-channel gyroscope, one-button takeoff and landing, ascent and descent, forward and backward movement, left and right side flight, turning, headless mode, optical flow positioning hovering, and supports remote electronically tunable infrared camera 103. It supports horizontal / overhead / vertical shooting, gesture photography, video recording, emergency stop, trajectory flight, gravity sensing, music, 50x zoom, and automatic photography. The wireless communication module 106 is installed inside the drone 100. It is a single module that enables communication between the smart chip 101 and the remote controller 107, receives operation commands sent by the remote controller 107, and sends captured photos or videos to the remote controller 107. The wireless communication module 106 provides wireless remote communication for this invention. Preferably, it is a multi-protocol communication module that supports Bluetooth, Zigbee, Thread, Proprietary, and Wi-Fi, with a maximum flash memory of 3200 / RAM of 512 (kB) and an output power range of -20 to 19.5 dBm.

[0036] The lifebuoy 200 provides the buoyancy required for rescue in this invention. Preferably, it is a professional lifebuoy conforming to the national standard GB4302-2008, orange, solid, made of PP plastic, 4.3KG adult size, with a load capacity of 280 catties, high-brightness reflective tape, suitable for nighttime rescue, and equipped with a high-strength nylon rope. It is CCS certified and suitable for use in open water. Preferably, it is an adult size (4.3KG): outer diameter 760mm, inner diameter 460mm, thickness 110mm. The larger size and weight provide stronger buoyancy, accommodating adults of different body types and needs. It provides stable buoyancy whether on a calm lake or a turbulent sea.

[0037] The rope 300 provides a flexible connection between the drone 100 and the lifebuoy 200 for this invention. Preferably, it is a high-strength nylon rope, certified by the classification society CCS, and suitable for use in open water.

[0038] The claw 400 provides a rigid connection between the drone 100 and the lifebuoy 200. Preferably, it is an electronically controlled aluminum alloy robotic claw that can open or close according to the instructions of the smart chip 101, thereby fixing or releasing the lifebuoy 200. Preferably, there are 3 or 4 claws. The minimum operating temperature is -30℃, the maximum operating temperature is 90℃, the minimum power supply voltage is 1V, the maximum power supply voltage is 9.5V, the width is 3.6mm, the height is 1.3mm, and the length is greater than the difference between the outer and inner diameters of the lifebuoy 200 or greater than 300mm, that is, the outer diameter of an adult (4.3KG) is 760mm minus the inner diameter.

[0039] like Figure 2 The diagram shows the working method of the intelligent lifebuoy drone in this invention.

[0040] Its characteristic is that it employs the following working steps: Step 1: The intelligent lifebuoy drone identifies the drowning victim's location and uses rope 300 to drag the lifebuoy 200 vertically towards the victim; Figure 3 The diagram shown illustrates step 1 of the intelligent lifebuoy drone in this invention.

[0041] Sub-step 1-1: Rescuers send rescue commands to the intelligent lifebuoy drone via remote controller 107; Sub-step 1-1-1: Rescuers activate the waterproof rotor 102 of the smart lifebuoy drone via remote control 107, and the smart lifebuoy drone takes off. Rescuers then activate the infrared camera 103 via remote control 107 to capture infrared images of the drowning victim at the scene. Furthermore, the operator can launch one or two intelligent lifebuoy drones around swimming pools, reservoirs, rivers, lakes, seas, and open water bodies, with several other intelligent lifebuoy drones on standby. Sub-step 1-1-2: Rescuers use remote control 107 to mark the location of the drowning victim with a red frame in the infrared on-site image of the drowning victim and send rescue instructions; Furthermore, only the smart lifebuoy drones that receive flight commands are powered on and operational, while the other smart lifebuoy drones are powered off and on standby. Sub-steps 1-2: The intelligent lifebuoy drone identifies the location of the drowning victim and flies towards them; Sub-step 1-2-1: The intelligent lifebuoy drone uses infrared image guidance to track the drowning person. The intelligent chip 101 controls the flight speed, altitude and direction of the intelligent lifebuoy drone by adjusting the waterproof rotor 102, and flies towards the position of the drowning person marked by the red box in the infrared on-site image. Furthermore, upon receiving the rescue command, the intelligent lifebuoy drone adjusts its flight speed, altitude, and direction, while simultaneously activating the infrared camera 103, microphone and speaker 104, and GPS positioning module 105; Furthermore, the infrared image guidance operation includes the following steps; Sub-step 1-2-1-1: The smart chip 101 activates the infrared camera 103 to capture infrared images of the drowning victim at the scene, and activates the microphone and speaker 104 to collect sounds from the drowning victim at the scene. Sub-step 1-2-1-2: The intelligent chip 101 runs a preprocessing algorithm to filter, convert the format, identify the target, and extract features from the infrared on-site image of the drowning victim captured by the infrared camera 103. Furthermore, the intelligent chip 101 reduces image noise through Gaussian filtering and preprocesses the real-time acquired infrared images of the drowning victim using the following formula:

[0042] in, It is a Gaussian kernel, and k is the kernel size. The image shows an infrared image of a drowning victim, where x and y are the pixels in the horizontal and vertical directions of the infrared image, respectively. Sub-steps 1-2-1-3: The smart chip 101 runs a preprocessing algorithm to filter, convert the format, identify the target, and extract features of the drowning victim's on-site sound collected by the microphone and speaker 104. In sub-step 1-2-1-4, the intelligent chip 101 processes the infrared on-site image frames of the drowning victim that have been preprocessed in sub-step 1-2-1-2. As input, the resolution is W H C (width W, height H, number of channels C) represents the infrared on-site image frames of the drowning victim as a time series. ,in It is the t-th frame; Furthermore, the intelligent chip 101 adjusts the size of the infrared image frames of the drowning victim to fit the model's input, using the following formula:

[0043] in These are the model input dimensions; Sub-steps 1-2-1-5: The intelligent chip 101 normalizes the pixel values ​​of the infrared on-site image frame of the drowning victim using the following formula:

[0044] in and These are the mean and standard deviation; In sub-step 1-2-1-6, the intelligent chip 101 takes the preprocessed drowning victim's on-site audio from sub-step 1-2-1-3 as input. The input drowning victim's on-site audio is... Representing the sounds of a drowning victim as a time series ,in It is the sound at time t; In sub-steps 1-2-1-7, the intelligent chip 101 extracts features from the infrared image frames of the drowning victim in sub-step 1-2-1-5 and the sounds of the drowning victim in sub-step 1-2-1-6. A convolutional neural network is used to extract the features of the infrared image frames and the sounds of the drowning victim. Here, the CNN is set to have L layers, and the output formula of the l-th layer is:

[0045] in It is the output of the l-th layer. It is the output of the ll-th layer. Convolutional kernel weights It's a bias. It is an activation function. Indicates the convolution operation; Sub-steps 1-2-1-8 output the feature map F of the infrared on-site image frame of the drowning victim, with a size of [missing information]. × × ,in , , These are the feature dimensions: width W, height H, and number of channels C. Sub-steps 1-2-1-9: The intelligent chip 101 performs drowning target detection based on the drowning victim's on-site sounds, calculates whether the feature map of the drowning victim's on-site sounds has the sound feature of calling for help, calculates the matching probability of the sound feature of calling for help, and sets it as the sound confidence level C1. Sub-step 1-2-1-10: The intelligent chip 101 performs drowning target detection based on the feature map F of the drowning victim's infrared on-site image frame, dividing the drowning victim's infrared on-site image frame into... A grid is generated, with each grid predicting B bounding boxes, and each bounding box containing center coordinates. Given the width w and height h, calculate whether the feature map of the infrared on-site image frame of the drowning victim has the image features of the drowning victim, calculate the probability of the drowning victim's existence, and set it as the image confidence level C2; Sub-step 1-2-1-11: Smart chip 101 calculates the category probability P(c|object), representing the probability that the target belongs to category C2. ,in It is the image confidence C2 category score; In sub-step 1-2-1-12, based on the sound confidence C1 from sub-step 1-2-1-9 and the image confidence C2 from sub-step 1-2-1-10, the intelligent chip 101 outputs the drowning victim target detection formula as follows: ; In sub-step 1-2-1-13, the intelligent chip 101 further calculates the total confidence C and marks the bounding box of the drowning victim in the infrared on-site image frame, which represents the probability that the bounding box contains the drowning victim target and the prediction is correct:

[0046] in It is the intersection-union ratio of the predicted bounding box and the ground truth bounding box, which can be obtained by statistical analysis based on historical data; Sub-step 1-2-1-14: The intelligent chip 101 performs non-maximum suppression to remove overlapping bounding boxes from the infrared on-site image frames of the drowning victim, retaining the boxes with the highest confidence, and calculating the intersection-union ratio (IUU) using the following formula:

[0047] ,if If true, then remove The bounding box retains the confidence level. The threshold value is 0.5. Sub-step 1-2-1-15: The intelligent chip 101 marks the drowning target detected in the infrared on-site image frame of the drowning victim with a red bounding box. The coordinates of the bounding box are: Draw a rectangle; Sub-step 1-2-1-17: The smart chip 101 adds a type label and confidence level above the bounding box of the drowning victim in the infrared on-site image frame of the drowning victim; In sub-step 1-2-1-18, the intelligent chip 101 estimates the distance from the intelligent lifebuoy drone to the drowning person based on the shape and size of the drowning person marked in red in the infrared live image and the human body size in historical infrared live image data; and determines whether the distance estimation meets the error condition based on the drowning person target in sub-step 1-2-1-15. If it does, the coordinates of the drowning person are located. ,in It is the estimated three-dimensional coordinates of the drowning victim in longitude, latitude, and altitude; Sub-step 1-2-1-19: The smart chip 101 locates the coordinates of the drowning victim according to step 1-2-1-18. Step 1-2-1-15: Real-time image acquisition Output the bounding box of the drowning person To accurately locate a drowning victim, the monocular ranging formula is as follows:

[0048] in, It is the actual height of the drowning victim. is the actual altitude of the intelligent lifebuoy drone, and f is the focal length of the camera; Sub-step 1-2-1-20: The intelligent chip 101 calculates the relative coordinates of the intelligent lifebuoy drone using the following formula:

[0049] in, , These are the optical center coordinates in longitude and latitude. and It refers to the focal length in terms of longitude and latitude; Sub-step 1-2-1-21: The intelligent chip 101 obtains the coordinates of the intelligent lifebuoy drone's location. The total Euclidean distance between the intelligent lifebuoy drone and the drowning victim is calculated using the following formula:

[0050] Sub-step 1-2-1-22: The intelligent chip (101) performs a global path planning algorithm to select all points the intelligent lifebuoy drone passes through to the drowning person. The formula for calculating the actual cost of the intelligent lifebuoy drone flying from the starting position to the drowning person is as follows:

[0051] in, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-23, based on the Euclidean total distance between the intelligent lifebuoy drone and the drowning person in sub-step 1-2-1-21. The actual cost in sub-step 1-2-1-22 The intelligent chip 101 further obtained the cost function for the intelligent lifebuoy drone to fly to the location of the drowning person as follows:

[0052] Sub-step 1-2-1-24: The elected intelligent lifebuoy drone samples the nodes of each path that can reach the drowning person, gradually constructing a tree structure, and the optimized path formula is:

[0053] in, It is the smoothing coefficient. Obstacle avoidance loss, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-25: The intelligent chip 101 performs distributed information sharing, and each intelligent lifebuoy drone broadcasts its own status, including the distance between it and the drowning person, the percentage of battery power, and the real-time location collected by the GPS positioning module 105, and uses time division multiple access to prevent channel congestion. In sub-step 1-2-1-26, the intelligent chip 101 elects the closest intelligent lifebuoy drone, meaning all intelligent lifebuoy drones elect the closest one to the drowning victim for rescue. Furthermore, since the positions of all intelligent lifebuoy drones are constantly changing, the primary strategy for dynamic election is to select the intelligent lifebuoy drone with the smallest distance to the drowning victim. A secondary strategy is to select the intelligent lifebuoy drone with the highest battery percentage if the distance difference between multiple intelligent lifebuoy drones is less than 0.5m. The selected intelligent lifebuoy drone then broadcasts confirmation, while the other intelligent lifebuoy drones switch to observation mode. Sub-step 1-2-1-27: The selected intelligent lifebuoy drone needs to construct a local map. Based on the GPS positioning module 105 and the infrared camera 103, a local map is constructed and the drowning person is accurately located on the map. Sub-step 1-2-1-28: The intelligent chip 101 performs motion control on the flight of the intelligent lifebuoy drone, and the PID output of the waterproof rotor 102... The control command formula is: ; in For time variables, Represents proportional, integral, and derivative gain, and output error. Locating the coordinates of a drowning victim Coordinates of the location of the intelligent lifebuoy drone .

[0054] Sub-step 1-2-1-29: The intelligent chip 101 predicts and controls the flight path of the intelligent lifebuoy drone to the drowning victim, and the formula for selecting the optimal path is: ; in, It is a predicted future state. It controls the input sequence; In sub-step 1-2-1-30, if the distance to the drowning victim does not meet the preset threshold, return to sub-step 1-2-1-18 to re-obtain the drowning victim's coordinates. Step 1-2-1-21: Reacquire the coordinates of the intelligent lifebuoy drone's current location. This is to prevent the drowning person's position from changing during the rescue process; Sub-step 1-2-1-31: The smart chip 101 performs anomaly monitoring. If the drowning target is lost for more than 3 seconds, a re-election is triggered, and the process returns to sub-step 1-2-1-1. Sub-step 1-2-1-32: If the smart chip 101 detects that the smart lifebuoy drone's battery percentage is below 20%, it will return to home and request the remote controller 107 to replace the smart lifebuoy drone. Sub-step 1-2-2: The intelligent lifebuoy drone continuously captures infrared images of the drowning victim and adjusts the angle of the infrared camera 103 to update the position of the drowning victim marked in red in the infrared images, ensuring that the position of the drowning victim marked in red is located in the center of the infrared images. Furthermore, the smart chip 101 acquires the images captured by the infrared camera 103, the sound captured by the microphone and speaker 104, and the real-time location captured by the GPS positioning module 105, and sends them to the remote control 107 via the wireless communication module 106. Sub-steps 1-2-3: Rescuers can use the intelligent lifebuoy drone as a rescue drone and start another intelligent lifebuoy drone as an observation drone via remote controller 107. Furthermore, the observation drone continuously captures infrared images of the rescue drone and the drowning victim, updates the position of the drowning victim marked in red in the infrared images, and the observation drone can correct the position, direction and speed of the rescue drone through wireless communication module 106. Furthermore, if there are multiple drowning victims, the operator can use the remote control 107 to directly control multiple smart lifebuoy drones to fly to each drowning victim and carry out rescue operations on multiple drowning victims at the same time. Sub-steps 1-3: When close to the drowning person, the intelligent lifebuoy drone controls the release of the claw buckle 400 to lower the lifebuoy 200, and only uses the rope 300 to drag the upper part of the lifebuoy 200 to ensure that the lifebuoy 200 is in a vertical position; Sub-step 1-3-1: The intelligent lifebuoy drone estimates the distance between itself and the drowning person based on the shape and size of the drowning person marked in red in the infrared live image and the human body size in historical infrared live image data. Preferably, the remaining flight time between the two is used to estimate the distance between them. The intelligent lifebuoy drone estimates the remaining flight time based on the rate of change of the shape and size of the drowning person marked in red in the infrared live image and the time already flown. Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy 200 as it approaches the drowning person, ensuring that the drowning person's head is located in the center of the infrared image. Sub-step 1-3-2: If the intelligent chip 101 estimates that the remaining flight time is short, the waterproof rotor 102 is adjusted to gradually reduce the flight speed and altitude. Preferably, when the intelligent chip 101 estimates that the remaining flight time is 1-2 minutes, it is closer to the drowning person. Further, gradually reducing the flight speed needs to ensure that when the drone flies to the drowning person, its flight speed can drop to 0 and hover above the drowning person so that the drowning person can grab the lifebuoy 200. Further, gradually reducing the flight altitude needs to ensure that when the lifebuoy 200 is lowered, the lower part of the lifebuoy 200 can touch the water surface to buffer the falling speed of the lifebuoy 200 and prevent the lifebuoy 200 from falling too fast, which would pull the intelligent lifebuoy drone through the rope 300 and excessively interfere with the normal flight of the intelligent lifebuoy drone. Sub-step 1-3-3: When the intelligent chip 101 estimates that the flight altitude reduction meets the requirements, it controls the release of the claw buckle 400 to lower the lifebuoy 200. The drone 100 and the lifebuoy 200 separate, the lifebuoy 200 falls and its lower part touches the water surface, and the upper part of the lifebuoy 200 is kept connected to the drone 100 by the rope 300. Sub-steps 1-4: The intelligent lifebuoy drone identifies the drowning victim's face and controls the rope 300 to drag the lifebuoy 200 in a vertical position towards the drowning victim's face and head. Sub-step 1-4-1: The smart chip 101 captures infrared images of the drowning victim using the infrared camera 103 and identifies the drowning victim's face; Sub-step 1-4-2: The intelligent chip 101 determines whether the current flight direction of the intelligent lifebuoy drone is aligned with the face and head of the drowning person; if so, it continues to fly in the direction of the face and head of the drowning person; if not, the intelligent chip 101 adjusts the waterproof rotor 102 to change the current direction to fly in the direction of the face and head of the drowning person. Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the face and head of the drowning person and ensures that the face and head of the drowning person are located in the center of the infrared live image. Step 2: The intelligent lifebuoy drone identifies the drowning victim's face and drags rope 300 in the opposite direction to the victim's face, controlling the lifebuoy 200 to rotate downwards around the victim's head and secure it. Figure 4 The diagram shown illustrates step 2 of the intelligent lifebuoy drone in this invention.

[0055] Sub-step 2-1: The intelligent lifebuoy drone hovers above the drowning person, allowing the drowning person to grab the lifebuoy 200, and identifies whether the drowning person has grabbed the lower part of the closest lifebuoy 200; Sub-step 2-1-1: The intelligent chip 101 captures an infrared image of the drowning victim using the infrared camera 103 and identifies whether the drowning victim can touch the lifebuoy 200. If so, it maintains its current position and hovers above the drowning victim. If not, the intelligent chip 101 adjusts the waterproof rotor 102 to control the lifebuoy 200 to approach the drowning victim's face at an extremely low flight speed. Further, the extremely low flight speed is no more than 1 meter per second. Further, if wind interferes with the rope 300 dragging the lifebuoy 200, causing the lifebuoy 200 to rotate or deviate from the drowning victim's face, the intelligent chip 101 adjusts the waterproof rotor 102 to change its current speed and direction, flying towards the drowning victim's face and head. Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy 200 as it approaches the face of the drowning person, ensuring that the drowning person's face is located in the center of the infrared live image. Sub-step 2-1-2: The smart chip 101 obtains the current GPS coordinates of the drowning person through the GPS positioning module 105 and reports the current GPS coordinates of the drowning person to the remote control 107 through the wireless communication module 106; the smart chip 101 picks up the sound at the scene through the microphone and speaker 104, or plays voice messages to the scene to comfort the drowning person and guide the drowning person to use the lifebuoy correctly. Sub-step 2-1-3: Since the lifebuoy 200 is vertically positioned close to the water surface, the drowning person can only grab the lower part of the lifebuoy 200 and allow their head and hands to naturally enter the holes of the lifebuoy 200; the smart chip 101 captures an infrared image of the drowning person using the infrared camera 103 and identifies whether the drowning person has already grabbed the lower part of the lifebuoy 200, or whether the drowning person's head and hands have naturally entered the holes of the lifebuoy 200. If so, proceed to sub-step 2-2; otherwise, proceed to sub-step 2-1-1; furthermore, if there is wind interference at the drowning person's location causing the rope 300 to drag the lifebuoy 200, resulting in the lifebuoy 200 rotating or deviating from the drowning person's face, once the drowning person has grabbed the lower part of the lifebuoy 200, the rotation or deviation of the lifebuoy 200 will disappear on its own, and the smart chip 101 will not need to adjust its flight speed and direction; Sub-step 2-2: If the drowning person manages to grab the lower part of the nearest lifebuoy 200, the intelligent lifebuoy drone identifies the drowning person's face position and flies behind the drowning person. It controls the rope 300 to drag the upper part of the lifebuoy 200 to flip it to a horizontal position behind the drowning person, and the lifebuoy 200 covers the drowning person from head to toe. Sub-step 2-2-1: The intelligent chip 101 captures infrared images of the drowning victim using the infrared camera 103, identifies the victim's face position, adjusts the waterproof rotor 102 to fly at an extremely low speed behind the drowning victim, and controls the rope 300 to drag the upper part of the lifebuoy 200 to flip behind the drowning victim; further, the intelligent chip 101 adjusts the waterproof rotor 102 to control the intelligent lifebuoy drone to flip and fly with the lower part of the lifebuoy 200 as the center and the sum of the diameter of the lifebuoy 200 and the length of the rope 300 as the radius; Sub-step 2-2-2: The smart chip 101 captures an infrared image of the drowning victim using the infrared camera 103, identifies the position of the lifebuoy 200 and the drowning victim's head, and determines whether the lifebuoy 200 has flipped over the drowning victim's head. If so, the lifebuoy has successfully wrapped around the drowning victim, and the process proceeds to sub-step 2-3. If not, the lifebuoy has not successfully wrapped around the drowning victim, and the process proceeds to sub-step 2-2-1 to continue the flipping operation.

[0056] Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy 200 as it rolls around the drowning person, ensuring that the drowning person's head is in the center of the infrared image. Sub-steps 2-3: The intelligent lifebuoy drone identifies whether the lifebuoy 200 is horizontal and whether the drowning person's face and hands are on the lifebuoy 200, confirming that the lifebuoy 200 has successfully encased the drowning person; The intelligent chip 101 captures infrared images of the drowning victim using the infrared camera 103, and identifies the lifebuoy 200 and the drowning victim's head and hands. It determines whether the drowning victim's head and hands are on the lifebuoy 200, whether the image of the lifebuoy 200 is nearly a perfect circle, and whether the connection between the lifebuoy 200 and the rope 300 is on the drowning victim's back. If so, the lifebuoy 200 is horizontal and has successfully encircled the drowning victim. The rope 300 pulls the lifebuoy 200 from the drowning victim's back, and the process proceeds to step 3. If not, the drowning victim has not been successfully encircled, and the process proceeds to sub-step 2-2-1 to continue the flipping operation.

[0057] Step 3: The intelligent lifebuoy drone uses rope 300 and lifebuoy 200 to pull the trapped drowning person back to shore in a supine position. (For example...) Figure 5 The diagram shown illustrates step 3 of the intelligent lifebuoy drone in this invention.

[0058] Sub-step 3-1: The intelligent lifebuoy drone identifies and returns to the shore location; Sub-step 3-1-1: Rescuers send the GPS coordinates of the return-to-shore position to the intelligent lifebuoy drone via remote controller 107; Sub-step 3-1-2: The intelligent lifebuoy drone acquires and identifies the GPS coordinates of the return shore location; Sub-step 3-2: The intelligent lifebuoy drone plans the optimal flight route; Sub-step 3-2-1: The intelligent lifebuoy drone obtains the GPS coordinates of its current location through the GPS positioning module 105; Sub-step 3-2-2: The intelligent lifebuoy drone plans the optimal flight route based on the GPS coordinates of the return shore location, the current location, and the water imagery and obstacle conditions between the two, so that the drowning victim can avoid dangers and obstacles on the return journey. Preferably, the intelligent lifebuoy drone uses infrared image guidance to plan its path, referring to sub-steps 1-2-1 and 1-2-2, and monitors the drowning person's condition throughout the process, ensuring that the drowning person's head is located in the center of the infrared image. Sub-step 3-3: The intelligent lifebuoy drone follows the optimal flight path, dragging rope 300 and lifebuoy 200 to pull the trapped drowning person back to shore, keeping lifebuoy 200 horizontal throughout the process.

[0059] Sub-step 3-3-1: The intelligent lifebuoy drone flies along the optimal flight path, ensuring that the rope 300 pulls the lifebuoy 200 from the drowning person's back, and the drowning person returns to the shore with their back facing the shore in a backstroke position; the intelligent lifebuoy drone monitors the drowning person's condition throughout the process to ensure minimal water inhalation and maximum safety, with rescuers and the intelligent lifebuoy drone working together to ensure the safety of the drowning person's return route.

[0060] Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the drowning person's status and ensures that the drowning person's head is located in the center of the infrared on-site image. Sub-step 3-3-2: The rescue is completed. The intelligent lifebuoy drone returns to the standby point, awaits new instructions, and uses the claw buckle 400 to re-secure the lifebuoy 200.

Claims

1. A smart lifebuoy drone, characterized in that, The system includes a drone (100) and a lifebuoy (200), connected by a rope (300) and a claw buckle (400). The drone (100) includes a smart chip (101), a waterproof rotor (102), an infrared camera (103), a microphone and speaker (104), a GPS positioning module (105), a wireless communication module (106), and a remote controller (107). The system operates using the following steps: Step 1: The intelligent lifebuoy drone identifies the location of the drowning person and uses a rope (300) to drag the lifebuoy (200) to approach the drowning person in a vertical position; Step 2: The intelligent lifebuoy drone identifies the drowning victim's face position and drags the rope (300) in the opposite direction to the drowning victim's face position, controlling the lifebuoy (200) to flip down around the drowning victim's head and cover the drowning victim; Step 3: The intelligent lifebuoy drone drags the rope (300) and lifebuoy (200) to pull the trapped drowning person back to shore in a supine position.

2. The UAV according to claim 1, characterized in that, Step 1 includes the following sub-steps: Sub-step 1-1: Rescuers send rescue instructions to the intelligent lifebuoy drone via remote control (107); Sub-steps 1-2: The intelligent lifebuoy drone identifies the location of the drowning victim and flies towards them; Sub-steps 1-3: When close to the drowning person, the intelligent lifebuoy drone controls the release of the claw buckle (400) and lowers the lifebuoy (200), using only the rope (300) to drag the upper part of the lifebuoy (200) to ensure that the lifebuoy (200) is in a vertical position; Sub-steps 1-4: The intelligent lifebuoy drone identifies the drowning victim's face and controls the rope (300) to drag the lifebuoy (200) in a vertical position towards the drowning victim's face and head.

3. The UAV according to claim 2, characterized in that, Step 1-1 includes the following sub-steps: Sub-step 1-1-1: Rescuers activate the waterproof rotor (102) of the smart lifebuoy drone via remote control (107), and the smart lifebuoy drone takes off. Rescuers activate the infrared camera (103) via remote control (107) to capture infrared images of the drowning victim. Sub-step 1-1-2: Rescuers use a remote control (107) to mark the location of the drowning victim in the infrared on-site image with a red frame and send rescue instructions; Steps 1-2 include the following sub-steps: Sub-step 1-2-1: The intelligent lifebuoy drone uses infrared image guidance to track the drowning person. The intelligent chip (101) controls the flight speed, altitude and direction of the intelligent lifebuoy drone by adjusting the waterproof rotor (102) and flies towards the position of the drowning person marked in red in the infrared on-site image. Upon receiving the rescue order, the intelligent lifebuoy drone adjusts its flight speed, altitude, and direction, and simultaneously activates its infrared camera (103), microphone and speaker (104), and GPS positioning module (105). Sub-step 1-2-2: The intelligent lifebuoy drone continuously captures infrared images of the drowning victim and adjusts the angle of the infrared camera (103) to update the position of the drowning victim marked in red in the infrared images, ensuring that the position of the drowning victim marked in red is located in the middle of the infrared images; The smart chip (101) acquires the image captured by the infrared camera (103), the sound captured by the microphone and speaker (104), and the real-time location captured by the GPS positioning module (105), and sends them to the remote control (107) through the wireless communication module (106). Sub-steps 1-2-3: Rescuers can use the intelligent lifebuoy drone as a rescue drone and start another intelligent lifebuoy drone as an observation drone via remote control (107); the observation drone continuously captures infrared images of the rescue drone and the drowning victim, updates the position of the drowning victim marked in red in the infrared images, and the observation drone can correct the position, direction and speed of the rescue drone through the wireless communication module (106); If there are multiple drowning victims, the operator can use the remote control (107) to directly operate multiple smart lifebuoy drones to fly to each drowning victim and carry out rescue operations on multiple drowning victims at the same time.

4. The UAV according to claim 3, characterized in that, In step 1-2-1, the operation of infrared image guidance includes the following steps: Sub-step 1-2-1-1: The smart chip (101) turns on the infrared camera (103) to capture infrared images of the drowning victim, and turns on the microphone and speaker (104) to collect the sound of the drowning victim at the scene. Sub-step 1-2-1-2: The smart chip (101) runs a preprocessing algorithm to filter, convert the format, identify the target and extract features from the infrared on-site images of the drowning victim captured by the infrared camera (103); The intelligent chip (101) reduces image noise through Gaussian filtering and preprocesses the real-time acquired infrared images of drowning victims using the following formula: ； in, It is a Gaussian kernel, and k is the kernel size. The image shows an infrared image of a drowning victim, where x and y are the pixels in the horizontal and vertical directions of the infrared image, respectively. Sub-steps 1-2-1-3: The smart chip (101) runs a preprocessing algorithm to filter, convert the format, identify the target and extract features of the drowning victim’s on-site sound collected by the microphone and speaker (104); Sub-step 1-2-1-4: The intelligent chip (101) processes the infrared on-site image frames of the drowning victim that have been preprocessed in sub-step 1-2-1-2. As input, the resolution is W H C, representing width W, height H, and number of channels C, represents the infrared on-site image frames of the drowning victim as a time series. ,in It is the t-th frame; The intelligent chip (101) adjusts the size of the infrared on-site image frames of the drowning victim to fit the model's input, using the following formula: ； in These are the model input dimensions; Sub-steps 1-2-1-5: The intelligent chip (101) normalizes the pixel values ​​of the infrared on-site image frame of the drowning victim using the following formula: ； in and These are the mean and standard deviation; In sub-step 1-2-1-6, the intelligent chip (101) takes the preprocessed drowning victim sound from sub-step 1-2-1-3 as input. The input drowning victim sound is... Representing the sounds of a drowning victim as a time series ,in It is the sound at time t; In sub-step 1-2-1-7, the intelligent chip (101) extracts features from the infrared image frames of the drowning victim in sub-step 1-2-1-5 and the sounds of the drowning victim in sub-step 1-2-1-6. A convolutional neural network is used to extract the features of the infrared image frames and the sounds of the drowning victim. Here, the CNN is set to have L layers, and the output formula of the l-th layer is: ； in It is the output of the l-th layer. It is the output of the ll-th layer. Convolutional kernel weights It's a bias. It is an activation function. Indicates the convolution operation; Sub-steps 1-2-1-8 output the feature map F of the infrared on-site image frame of the drowning victim, with a size of [missing information]. × × ,in , , These are the feature dimensions: width W, height H, and number of channels C. Sub-steps 1-2-1-9: The smart chip (101) performs drowning target detection based on the drowning victim's on-site sound, calculates whether the feature map of the drowning victim's on-site sound has the sound feature of calling for help, calculates the matching probability of the sound feature of calling for help, and sets it as the sound confidence level C1. Sub-step 1-2-1-10: The intelligent chip (101) performs drowning target detection based on the feature map F of the drowning victim's infrared on-site image frame, dividing the drowning victim's infrared on-site image frame into... A grid is generated, with each grid predicting B bounding boxes, and each bounding box containing center coordinates. Given the width w and height h, calculate whether the feature map of the infrared on-site image frame of the drowning victim has the image features of the drowning victim, calculate the probability of the drowning victim's existence, and set it as the image confidence level C2; Sub-step 1-2-1-11, the intelligent chip (101) calculates the category probability P(c|object), which represents the probability that the target belongs to category C2. ,in It is the image confidence C2 category score; In sub-step 1-2-1-12, based on the sound confidence C1 in sub-step 1-2-1-9 and the image confidence C2 in sub-step 1-2-1-10, the intelligent chip (101) outputs the drowning victim target detection formula as follows: ; In sub-step 1-2-1-13, the intelligent chip (101) further calculates the total confidence C and marks the bounding box of the drowning victim in the infrared scene image frame, which represents the probability that the bounding box contains the drowning victim target and the prediction is correct: ； in It is the intersection-union ratio of the predicted bounding box and the ground truth bounding box, which can be obtained by statistical analysis based on historical data; Sub-step 1-2-1-14: The intelligent chip (101) performs non-maximum suppression to remove overlapping bounding boxes in the infrared on-site image frames of the drowning victim, retaining the boxes with the highest confidence, and calculating the intersection-union ratio (IUU) using the following formula: ； ,if If true, then remove The bounding box retains the confidence level; The threshold value is 0.

5. Sub-step 1-2-1-15: The intelligent chip (101) marks the drowning target detected in the infrared on-site image frame of the drowning victim with a red bounding box. The coordinates of the bounding box are: Draw a rectangle; Sub-step 1-2-1-17, the smart chip (101) adds a type label and confidence level above the bounding box of the drowning victim in the infrared on-site image frame of the drowning victim; Sub-step 1-2-1-18: The intelligent chip (101) estimates the distance from the intelligent lifebuoy drone to the drowning person based on the shape and size of the drowning person marked in red in the infrared live image and the size of the human body in the historical data of the infrared live image; and determines whether the distance estimation meets the error condition based on the drowning person target in sub-step 1-2-1-15. If it does, the coordinates of the drowning person are located. ,in It is the estimated three-dimensional coordinates of the drowning victim in longitude, latitude, and altitude; Sub-step 1-2-1-19: The smart chip (101) locates the coordinates of the drowning victim according to step 1-2-1-18. Step 1-2-1-15: Real-time image acquisition Output the bounding box of the drowning person To accurately locate a drowning victim, the monocular ranging formula is as follows: ； in, It is the actual height of the drowning victim. is the actual altitude of the intelligent lifebuoy drone, and f is the focal length of the camera; Sub-step 1-2-1-20: The intelligent chip (101) calculates the relative coordinates of the intelligent lifebuoy drone using the following formula: ； in, , These are the optical center coordinates in longitude and latitude. and It refers to the focal length in terms of longitude and latitude; Sub-step 1-2-1-21: The intelligent chip (101) obtains the coordinates of the location of the intelligent lifebuoy drone. The total Euclidean distance between the intelligent lifebuoy drone and the drowning victim is calculated using the following formula: ； Sub-step 1-2-1-22: The intelligent chip (101) performs a global path planning algorithm to select all points the intelligent lifebuoy drone passes through to the drowning person. The formula for calculating the actual cost of the intelligent lifebuoy drone flying from the starting position to the drowning person is as follows: ； in, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-23, based on the Euclidean total distance between the intelligent lifebuoy drone and the drowning person in sub-step 1-2-1-21. The actual cost in sub-step 1-2-1-22 The intelligent chip (101) further obtains the cost function for the intelligent lifebuoy drone to fly to the location of the drowning person as follows: ； Sub-step 1-2-1-24: The elected intelligent lifebuoy drone samples the nodes of each path that can reach the drowning person, gradually constructing a tree structure, and the optimized path formula is: ； in, It is the smoothing coefficient. Obstacle avoidance loss, The distance between two adjacent points is represented by , and m represents all points traversed by the intelligent lifebuoy drone. ; Sub-step 1-2-1-25, the smart chip (101) performs distributed information sharing, each smart lifebuoy drone broadcasts its own status, including the distance between it and the drowning person, the percentage of battery power and the real-time location collected by the GPS positioning module (105), and uses time division multiple access to prevent channel congestion; Sub-step 1-2-1-26: The smart chip (101) elects the nearest smart lifebuoy drone, that is, all smart lifebuoy drones elect the smart lifebuoy drone closest to the drowning person for rescue. All the smart lifebuoy drones are constantly changing positions. The primary strategy of dynamic election is to select the smart lifebuoy drone that is closest to the drowning person. The secondary strategy is to select the smart lifebuoy drone with the highest battery percentage if the distance difference between multiple smart lifebuoy drones is less than a certain value. The selected smart lifebuoy drone broadcasts confirmation, while the other smart lifebuoy drones switch to observation mode. Sub-step 1-2-1-27: The selected intelligent lifebuoy drone needs to construct a local map. Based on the GPS positioning module (105) and the infrared camera (103), a local map is constructed and the drowning person is accurately located on the map. Sub-step 1-2-1-28: The intelligent chip (101) performs motion control on the flight of the intelligent lifebuoy drone, and the PID output of the waterproof rotor (102) The control command formula is: ； in For time variables, Represents proportional, integral, and derivative gain, and output error. Locating the coordinates of a drowning victim Coordinates of the location of the intelligent lifebuoy drone ; Sub-step 1-2-1-29: The intelligent chip (101) predicts and controls the flight path of the intelligent lifebuoy drone to the drowning victim, and the optimal path selection formula is: ； in, It is a predicted future state. It controls the input sequence; In sub-step 1-2-1-30, if the distance to the drowning victim does not meet the preset threshold, return to sub-step 1-2-1-18 to re-obtain the drowning victim's coordinates. Step 1-2-1-21: Reacquire the coordinates of the intelligent lifebuoy drone's current location. This is to prevent the drowning person's position from changing during the rescue process; Sub-step 1-2-1-31: The smart chip (101) performs anomaly monitoring. If the drowning target is lost for more than a certain period of time, a re-election is triggered, and the process returns to sub-step 1-2-1-1. In sub-step 1-2-1-32, if the smart chip (101) detects that the battery percentage of the smart lifebuoy drone is lower than a certain value, it will return to home and request the remote controller (107) to replace the smart lifebuoy drone.

5. The UAV according to claim 3, characterized in that, Steps 1-3 include the following sub-steps: Sub-step 1-3-1: The intelligent lifebuoy drone estimates the distance between itself and the drowning person based on the shape and size of the drowning person marked in red in the infrared live image and the human body size in historical infrared live image data. Preferably, the remaining flight time between the two is used to estimate the distance between them. The intelligent lifebuoy drone estimates the remaining flight time based on the rate of change of the shape and size of the drowning person marked in red in the infrared live image and the time already flown. The intelligent lifebuoy drone uses infrared image guidance to track the lifebuoy (200) as it approaches the drowning person; Sub-step 1-3-2: If the intelligent chip (101) estimates that the remaining flight time is short, it will adjust the waterproof rotor (102) to gradually reduce the flight speed and altitude; Gradually reducing the flight speed requires ensuring that when the drone reaches the drowning person, its flight speed can be reduced to 0, hovering above the drowning person so that the drowning person can grab the lifebuoy (200). Gradually reducing the flight altitude requires ensuring that when the lifebuoy (200) is lowered, the lower part of the lifebuoy (200) can touch the water surface to buffer the falling speed of the lifebuoy (200) and avoid the lifebuoy (200) falling too fast, which would pull the smart lifebuoy drone through the rope (300) and excessively interfere with the normal flight of the smart lifebuoy drone. Sub-step 1-3-3: When the intelligent chip (101) estimates that the flight altitude reduction meets the requirements, it controls the release of the claw buckle (400) to lower the life ring (200). The drone (100) and the life ring (200) separate, the life ring (200) falls and its lower part touches the water surface, and the upper part of the life ring (200) is connected to the drone (100) by the rope (300).

6. The UAV according to claim 1, characterized in that, Steps 1-4 include the following sub-steps: Sub-step 1-4-1: The smart chip (101) captures infrared images of the drowning victim through the infrared camera (103) and identifies the drowning victim's face; Sub-step 1-4-2: The smart chip (101) determines whether the current flight direction of the smart lifebuoy drone is aligned with the face and head of the drowning person; if so, it continues to fly with the current direction aligned with the face and head of the drowning person; if not, the smart chip (101) adjusts the waterproof rotor (102) to change the current direction to fly with the face and head of the drowning person. Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the face and head of the drowning person, and ensures that the face and head of the drowning person are located in the center of the infrared on-site image.

7. The UAV according to any one of claims 1 to 6, characterized in that, Step 2 includes the following sub-steps: Sub-step 2-1: The intelligent lifebuoy drone hovers above the drowning person, allowing the drowning person to grab the lifebuoy (200), and identifies whether the drowning person has grabbed the lower part of the nearest lifebuoy (200); Sub-step 2-2: If the drowning person manages to grab the lower part of the nearest lifebuoy (200), the intelligent lifebuoy drone identifies the drowning person's face position and flies behind the drowning person. It controls the rope (300) to drag the upper part of the lifebuoy (200) to flip it to a horizontal position behind the drowning person, and the lifebuoy (200) wraps around the drowning person from head to toe. Sub-steps 2-3: The intelligent lifebuoy drone identifies whether the lifebuoy (200) is horizontal and whether the drowning person's face and hands are on the lifebuoy (200), confirming that the lifebuoy (200) has successfully encased the drowning person.

8. The UAV according to claim 7, characterized in that, Step 2-1 includes the following sub-steps: Sub-step 2-1-1: The smart chip (101) captures an infrared image of the drowning victim using an infrared camera (103) and identifies whether the drowning victim can touch the lifebuoy (200). If so, it keeps the current position hovering above the drowning victim. If not, the smart chip (101) adjusts the waterproof rotor (102) to control the lifebuoy (200) to approach the drowning victim's face at an extremely low flight speed. Furthermore, if there is wind interference at the drowning victim's location that causes the rope (300) to drag the lifebuoy (200), resulting in the lifebuoy (200) rotating or deviating from the drowning victim's face, the smart chip (101) adjusts the waterproof rotor (102) to change the current speed and direction and fly towards the drowning victim's face and head. Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy (200) as it approaches the drowning person's face, and ensures that the drowning person's face is located in the center of the infrared live image; Sub-step 2-1-2: The smart chip (101) obtains the current GPS coordinates of the drowning person through the GPS positioning module (105) and reports the current GPS coordinates of the drowning person to the remote controller (107) through the wireless communication module (106); the smart chip (101) picks up the sound at the scene through the microphone and speaker (104) or plays voice messages to the scene to comfort the drowning person and guide the drowning person to use the lifebuoy correctly; Sub-step 2-1-3: Since the lifebuoy (200) is in a vertical position close to the water surface, the drowning person can only grab the lower part of the lifebuoy (200) so that their head and hands can naturally enter the hole of the lifebuoy (200); the smart chip (101) takes infrared images of the drowning person through the infrared camera (103) and identifies whether the drowning person has grabbed the lower part of the lifebuoy (200) or whether the drowning person's head and hands have naturally entered the hole of the lifebuoy (200). If so, proceed to sub-step 2-2; if not, proceed to sub-step 2-1-1; furthermore, if there is wind interference at the drowning person's location and the rope (300) drags the lifebuoy (200), causing the lifebuoy (200) to rotate or deviate from the drowning person's face, once the drowning person has grabbed the lower part of the lifebuoy (200), the rotation or deviation of the lifebuoy (200) will disappear by itself, and the smart chip (101) does not need to adjust the flight speed and direction; Step 2-2 includes the following sub-steps: Sub-step 2-2-1: The smart chip (101) captures infrared images of the drowning victim through the infrared camera (103), identifies the position of the drowning victim's face, adjusts the waterproof rotor (102) to fly to the back of the drowning victim at an extremely low flight speed, and controls the rope (300) to drag the upper part of the lifebuoy (200) to flip it to the back of the drowning victim. The smart chip (101) adjusts the waterproof rotor (102) and controls the smart lifebuoy drone to flip and fly with the lower part of the lifebuoy (200) as the center and the sum of the diameter of the lifebuoy (200) and the length of the rope (300) as the radius; Sub-step 2-2-2: The smart chip (101) captures infrared images of the drowning victim through the infrared camera (103), identifies the position of the lifebuoy (200) and the drowning victim's head, and determines whether the lifebuoy (200) has flipped over the drowning victim's head. If so, the lifebuoy has successfully wrapped around the drowning victim and proceeds to sub-step 2-3. If not, the lifebuoy has not successfully wrapped around the drowning victim and proceeds to sub-step 2-2-1 to continue the flipping operation. Preferably, the intelligent lifebuoy drone refers to sub-steps 1-2-1 and 1-2-2, and uses infrared image guidance to track the lifebuoy (200) as it rolls around the drowning person, ensuring that the drowning person's head is located in the center of the infrared image. In steps 2-3, the smart chip (101) captures an infrared image of the drowning victim using an infrared camera (103), identifies the lifebuoy (200) and the drowning victim's head and hands, determines whether the drowning victim's head and hands are on the lifebuoy (200), whether the image of the lifebuoy (200) is close to a perfect circle, and whether the connection between the lifebuoy (200) and the rope (300) is on the drowning victim's back. If so, the lifebuoy (200) is horizontal and has successfully encircled the drowning victim. The rope (300) pulls the lifebuoy (200) from the drowning victim's back and proceeds to step 3. If not, the drowning victim has not been successfully encircled and proceeds to sub-step 2-2-1 to continue the flipping operation.

9. The UAV according to claim 1, 2, 3, 4, 5, 6, or 8, characterized in that, Step 3 includes the following sub-steps: Sub-step 3-1: The intelligent lifebuoy drone identifies and returns to the shore location; Sub-step 3-2: The intelligent lifebuoy drone plans the optimal flight route; Sub-step 3-3: The intelligent lifebuoy drone drags the rope (300) and lifebuoy (200) along the optimal flight path to pull the trapped drowning person back to the shore, keeping the lifebuoy (200) horizontal throughout the process.

10. The UAV according to claim 9, characterized in that, Step 3-1 includes the following steps: Sub-step 3-1-1: Rescuers send the GPS coordinates of the return-to-shore position to the intelligent lifebuoy drone via remote control (107); Sub-step 3-1-2: The intelligent lifebuoy drone acquires and identifies the GPS coordinates of the return shore location; Step 3-2 includes the following steps: Sub-step 3-2-1: The intelligent lifebuoy drone obtains the GPS coordinates of its current location through the GPS positioning module (105); Sub-step 3-2-2: The intelligent lifebuoy drone plans the optimal flight route based on the GPS coordinates of the return shore location, the current location, and the water imagery and obstacle conditions between the two, so that the drowning victim can avoid dangers and obstacles on the return journey. The intelligent lifebuoy drone uses infrared image guidance for path planning and monitors the drowning victim's condition throughout the process. Step 3-3 includes the following steps: Sub-step 3-3-1: The intelligent lifebuoy drone flies along the optimal flight path, ensuring that the rope (300) pulls the lifebuoy (200) from the drowning person's back, and the drowning person returns to the shore with their back facing the shore in a backstroke position; the intelligent lifebuoy drone monitors the drowning person's condition throughout the process to ensure minimal water inhalation and maximum safety, and the rescue personnel and the intelligent lifebuoy drone work together to ensure the safety of the drowning person's return route; Preferably, the intelligent lifebuoy drone, referring to sub-steps 1-2-1 and 1-2-2, uses infrared image guidance to track the drowning person's status and ensures that the drowning person's head is located in the center of the infrared on-site image. Sub-step 3-3-2: The rescue is over. The smart lifebuoy drone returns to the standby point, waits for new instructions, and uses the claw buckle (400) to re-secure the lifebuoy (200).

Images