Collaborative operation system and method for unmanned aerial vehicle and airship

Abstract

The invention provides an unmanned aerial vehicle (UAV) collaborative operation system and a method thereof, and relates to the technical field of UAV collaborative operation. The system comprises an unmanned aerial vehicle, an unmanned ship and a ground station, the unmanned aerial vehicle comprises paddles used for providing power for the unmanned aerial vehicle and fixed to unmanned aerial vehicle arms through screws; an unmanned aerial vehicle-mounted wireless communication module; the depth vision camera is used for collecting obstacle information in a scene; the unmanned surface vehicle comprises an unmanned surface vehicle wireless communication module; the unmanned ship propeller is used for providing power for the unmanned ship; the ground station comprises a ground station wireless communication module; and the display screen module is used for displaying the digital map transmitted by the unmanned aerial vehicle. Through cooperative control of the unmanned aerial vehicle, the unmanned surface vehicle and the ground station, in combination with deep learning and an intelligent algorithm, efficient cooperative operation of polar region sea ice monitoring and the unmanned platform is realized.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) and unmanned surface vessel (USV) collaborative operation technology, and specifically to a UAV and USV collaborative operation system and method. Background Technology

[0002] In polar operations, such as marine resource exploration and environmental monitoring, unmanned systems are playing an increasingly important role in vast marine areas. Currently, most missions are carried out using standalone unmanned systems; however, this approach has some drawbacks: firstly, the payload capacity of a single platform is limited, and battery efficiency drops sharply under extreme low temperatures, making it difficult to simultaneously and continuously collect multi-dimensional data; secondly, the complex ice conditions in polar regions reduce navigation accuracy, and the reliability of individual obstacle avoidance systems is insufficient.

[0003] In order to promote the development of polar operations towards intelligence and efficiency, there is an urgent need for an efficient and flexible UAV / ship collaborative operation system and its corresponding collaborative operation methods that can meet the needs of base operations. Summary of the Invention

[0004] To address the technical problems of limited single-platform payload capacity, difficulty in completing multiple data acquisitions, decreased navigation accuracy, and difficulty in ensuring system obstacle avoidance in existing technologies, this invention proposes a collaborative operation system and method for unmanned aerial vehicles and ships.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: a UAV-boat collaborative operation system, the system comprising: a UAV, an unmanned surface vessel, and a ground station; The drone includes: propellers, which provide power to the drone and are fixed to the drone arm with screws; a drone-borne wireless communication module; and a depth vision camera, which is used to collect obstacle information in the scene. The unmanned surface vessel (USV) includes: a wireless communication module; and a propulsion system for providing power to the USV. The ground station includes: a ground station wireless communication module; and a display module for displaying the digital map transmitted by the drone. The system consists of a UAV (unmanned aerial vehicle) platform A, an unmanned surface vessel (USV) platform B, and a ground station platform C. Platform A constructs a digital map and identifies sea ice areas during operation. After completion, it transmits the data to platform C and then remains in standby mode. Platform C waits for the target position of platform B to be manually set before performing path planning and sending control commands. Platform B receives the control commands, moves to the target position according to the planned path, and then remains in standby mode. The path planning employs a Deep Q-Network (DQN) algorithm, which includes: initializing the main network, target network, and experience replay buffer; selecting actions according to an ε-greedy strategy during training; dynamically adjusting the strategy using immediate rewards from environmental feedback to maximize accumulated rewards for efficient obstacle avoidance; breaking data correlation through an experience replay mechanism; periodically updating the target network parameters; and finally outputting the planned path.

[0006] On the other hand, this application provides a method for the above-mentioned unmanned aerial vehicle (UAV) ship cooperative operation system, the method comprising: Step 1: System initialization, completing the parameter settings and function initialization of the UAV and unmanned surface vessel; Step 2: Read the initial position information according to the set initial position information; Step 3: The drone reads the target location information set by the ground station; Step 4: The drone reaches the target location using motion control algorithms; Step 5: Determine if the target location is the set target location. If the drone reaches the set target location, proceed to step 6; otherwise, proceed to step 3. Step 6: The drone reaches the set target position and hovers at a fixed altitude; Step 7: The depth vision camera carried by the drone rotates at a constant speed, and the camera captures images of objects in the current area; Step 8: Construct a digital map using the depth vision camera; Step 9: Identify sea ice in the image using a deep learning algorithm; Step 10: Mark the airworthiness area on the constructed digital map. After completion, proceed to Step 11. Step 11: Transmit map information from the drone to the ground station via wireless communication technology. After completion, proceed to step 18. Step 12: Determine whether the drone data transmission is complete. If complete, proceed to step 13; otherwise, proceed to step 11. Step 13, drone standby; Step 14: Set the destination position of the unmanned surface vessel via the ground station; Step 15: Set the unmanned surface vessel's movement path using a path planning algorithm; Step 16: Send ground station control commands to the unmanned surface vessel via wireless communication technology; Step 17: The unmanned surface vessel executes control commands and moves to the designated location; Step 18: Determine whether the unmanned surface vessel has reached the designated location. If it has, proceed to step 19; otherwise, proceed to step 16. Step 19: The system enters standby mode, waiting for new task instructions. The process ends upon completion.

[0007] Compared with the prior art, the beneficial effects of this application are: 1. This application overcomes the limitations of single-platform capabilities through multi-platform collaboration: Traditional single-UAV or single-UDV systems are limited by payload capacity, endurance, or sensor coverage, making it difficult to independently complete multi-data acquisition and obstacle avoidance tasks in complex sea areas. This application addresses this issue by using a UAV equipped with a depth camera and wireless communication module to acquire high-altitude sea ice images and construct digital maps; an UAV serving as the execution platform for operations in low-altitude or navigable surface areas; and a ground station acting as the central hub to coordinate path planning and control command issuance, thereby achieving multi-platform collaborative work.

[0008] 2. Deep learning-based sea ice identification improves monitoring accuracy: Traditional sea ice monitoring relies on manual visual interpretation or low-precision sensors, which struggles to cope with complex changes in lighting, climate, and sea ice morphology. This application employs a YOLO target detection algorithm model, optimizing network parameters through a training set to achieve rapid and accurate identification of sea ice targets (significantly improving detection accuracy and recall). After deployment, the model supports automatic identification of newly acquired images, and its inference speed meets the requirements of real-time monitoring.

[0009] 3. Intelligent path planning and obstacle avoidance ensure operational safety: In response to the complex polar environments, traditional path planning algorithms struggle to dynamically adapt to obstacle distribution and speed variations. This application employs 3D coordinate fusion: combining the 3D target position provided by the UAV's depth camera with the real-time position from the UAV's inertial navigation system, the relative positional relationship is calculated, thereby improving path planning accuracy.

[0010] DQN Algorithm Application: Deep Reinforcement Learning (DQN) is introduced to generate optimal path strategies through training, taking into account speed, heading, and obstacle avoidance. Compared to traditional AQN algorithms, this approach offers significant advantages. Or Dijkstra's algorithm is more adaptable.

[0011] Dynamic feedback adjustment: The propulsion system adjusts speed and heading commands in real time based on path planning results, inertial navigation and camera feedback data to ensure that the unmanned surface vessel smoothly avoids obstacles and accurately reaches the target location.

[0012] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0013] Figure 1 This is a structural diagram of the unmanned aerial vehicle (UAV) and boat collaborative operation system according to the present invention; Figure 2 This is a flowchart of the collaborative operation system based on unmanned aerial vehicles and boats according to the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0015] Unless the context requires otherwise, throughout the specification and claims, the word “comprising” and its variations, such as “including” and “having”, shall be understood to have an open, inclusive meaning, that is, to be interpreted as “including, but not limited to”.

[0016] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0017] The singular forms “a” and “the” used in this specification and the appended claims include plural references unless otherwise expressly stated herein. It should be noted that the term “or” is generally used to mean “and / or” unless otherwise expressly stated herein.

[0018] In the following description, in order to clearly demonstrate the structure and working method of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0019] The implementation details of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0020] This application proposes a sea ice identification system based on unmanned aerial vehicles (UAVs), and proposes a UAV-vessel collaborative control system and corresponding control methods, enabling UAVs and their onboard equipment to cooperate in the operation of the sea area.

[0021] This application is combined with Figure 1 This diagram illustrates the structure of an unmanned aerial vehicle (UAV) and unmanned surface vessel (USV) collaborative operation system, providing an explanation of the system. The core components of the system consist of three parts: the UAV, the USV, and the ground station. The system utilizes a dual-system platform (UAV and USV) and a ground station monitoring and control platform.

[0022] The drone includes: A1, the drone propeller, which powers the drone and is fixed to the drone arm with screws; A2, the drone's onboard wireless communication module, used for wireless information transmission; and A3, a depth vision camera, used to collect obstacle information in the scene. The depth vision camera (also known as a 3D camera) is a camera device capable of capturing distance information between objects and the camera, and generating a three-dimensional scene containing depth data.

[0023] The unmanned surface vessel (USV) includes: B1, which is the USV wireless communication module for wireless information transmission; and B2, which is the USV propulsion system for providing power to the USV.

[0024] The ground station includes: C1, a ground station wireless communication module for wireless information transmission; C2, a display module for displaying digital maps transmitted by the UAV; and C3, a ground station keyboard and mouse module for setting motion parameter information.

[0025] Specifically, during operations, the drone uses a depth vision camera to collect information about sea ice, obstacles, and other features on the sea surface. While the drone is collecting data above the sea level, the unmanned surface vessel (USV) is in its initial position. After acquiring a digital map and the location of the sea ice, the drone sends the map information to the ground station and then enters standby mode. The ground station sets the USV's destination and plans its path. The ground station then sends motion control commands to the USV via wireless communication. Once the USV reaches the designated location, it enters standby mode.

[0026] In some embodiments, this application provides a three-platform collaborative operation mode for an unmanned aerial vehicle (UAV) and unmanned surface vessel (USV) collaborative operation system. Specifically, the UAV has platform A, the USV has platform B, and the ground station has platform C. During operation, platform A constructs a digital map and identifies sea ice areas using a depth vision camera. While platform A acquires data information in the target area, platform B waits on the shore. After platform A completes the digital map construction, the data is transmitted to platform C via UAV-borne wireless communication technology. Platform A then enters standby mode. Platform C then waits for the target position of platform B to be manually set. Once the target position of platform B is acquired, platform C begins path planning and sends control commands to platform B via wireless communication technology. Simultaneously, upon receiving the control commands, platform B begins to move towards the designated target location area based on the planned path, and then enters standby mode. This cycle repeats continuously to achieve collaborative operation of the UAV and USV.

[0027] Furthermore, the optimal position calculation and path planning for the unmanned surface vessel are as follows: The ground station combines the target's position (3D coordinates provided by the depth vision camera) acquired by the UAV with the UAV's current position (position information provided by the UAV's inertial navigation system) to calculate the target's relative position to the UAV. Computer vision and path planning algorithms (DQN algorithm) are used to generate the optimal path for the UAV from its current position to the target position. Factors such as obstacles, speed, and heading are considered to generate a smooth path with obstacle avoidance. The processor calculates the heading adjustment commands the UAV should take based on the target position and heading information. Based on the target position, path planning results, inertial navigation system feedback, and camera feedback, the processor generates control commands for the UAV's propulsion system. These commands include: a speed command, generating the speed the UAV should achieve based on the target position and path planning; a heading command, calculating the angular velocity or turning angle required to adjust the heading based on the target direction; and a propulsion control command, where the propulsion system adjusts the thrust and direction of the thrusters to ensure the UAV moves towards the target position. By combining real-time feedback data from the inertial navigation system and camera, the UAV's heading and speed are adjusted to ensure precise mission execution.

[0028] In some embodiments, regarding the UAV / ship collaborative operation system of this application, the following will be combined with Figure 2 The workflow diagram is illustrated with a full-function setup, and the steps are as follows: Step 1: System initialization. Complete the parameter settings and function initialization of the UAV and unmanned surface vessel. After completion, proceed to Step 2. Step 2: If there is a set initial position information, read the position information, including the set flight altitude, latitude and longitude of the specified position, etc. After completion, proceed to Step 3. Step 3: The drone reads the target point location set by the ground station. After completion, proceed to step 4. Step 4: The drone uses motion control algorithms to reach the set position, and then proceeds to step 5. Step 5: Determine if the drone has reached the set location; otherwise, proceed to step 3. Step 6: Once the drone reaches the set position, hover at a fixed altitude. After completion, proceed to Step 7. Step 7: The depth vision camera carried by the drone rotates at a constant speed to capture objects in the current area. After completion, proceed to step 8. Step 8: Construct a digital map using an airborne depth vision camera. Once completed, proceed to Step 9. Step 9: Load the YOLO object detection model and its weight parameters that have been pre-trained on the sea ice dataset; The image of the sea ice area to be identified is obtained by acquiring the image or video frame to be detected through the acquisition device, and the original sea ice image is input into the recognition system; By performing preprocessing operations such as resizing, pixel normalization, and data format conversion on the input sea ice images, the images meet the input requirements of the YOLO object detection model and improve the accuracy and efficiency of subsequent recognition. By inputting the preprocessed sea ice image into the YOLO object detection model and performing forward inference, the raw detection results containing sea ice target candidate boxes, class probabilities, and confidence scores are obtained. By post-processing the target candidate boxes output by the YOLO object detection model, including filtering valid sea ice targets based on confidence thresholds and using the non-maximum suppression (NMS) algorithm to remove overlapping redundant boxes, accurate sea ice detection results are finally obtained. Among them, non-maximum suppression (NMS) is a core post-processing algorithm in computer vision tasks such as object detection and image segmentation, which is used to eliminate redundant detection boxes or segmentation regions and retain the most representative results. By drawing the bounding boxes, category labels, and confidence scores of detected sea ice targets on the original image, the sea ice detection results can be visualized. The result images can be displayed on the user interface or stored in a specified location as needed for subsequent analysis and application. By completing the above steps, the automatic recognition process of the input sea ice image is realized. The system enters a standby or end state, waiting for a new recognition task. After completion, it proceeds to step 10. Step 10: Mark the airworthiness area on the constructed digital map. After completion, proceed to Step 11. Step 11: Transmit map information from the drone to the ground station via wireless communication technology. After completion, proceed to step 12. Step 12: Determine whether the drone data transmission is complete. If complete, proceed to step 13; otherwise, proceed to step 11. Step 13: The drone is in standby mode. Once completed, proceed to step 14. Step 14: Set the destination of the unmanned surface vessel's movement via the ground station. After completion, proceed to Step 15. Step 15: Perform path planning and obstacle avoidance for the unmanned surface vessel using a path planning algorithm to obtain motion control information for the unmanned surface vessel. After completion, proceed to step 16. Step 16: Send ground station control commands to the unmanned surface vessel via wireless communication technology. After completion, proceed to step 17. Step 17: The unmanned surface vessel executes control commands and moves to the designated location. After completion, proceed to step 18. Step 18: Determine whether the unmanned surface vessel has reached the designated location. If it has, proceed to step 19; otherwise, proceed to step 16. Step 19: The system enters standby mode, waiting for new task instructions. The process ends upon completion.

[0029] In some embodiments, the YOLO object detection model in step 9 is trained based on a sea ice dataset. YOLO is a real-time object detection algorithm that employs an advanced neural network architecture and features fast detection speed and high accuracy. Its core idea is to treat the object detection task as a regression problem, thereby achieving end-to-end detection. The training process is as follows: Step 1. Collect raw image data containing sea ice distribution using camera equipment (such as drones or shore-based cameras) set up in the target area, ensuring that the collection process covers diverse scenarios with different climates, lighting, and sea ice morphology changes.

[0030] Step 2. Clean the raw sea ice images by removing blurry, noisy, duplicate or invalid image data, and select a dataset with high image quality, sufficient data volume and uniform distribution to improve the effectiveness and generalization ability of subsequent model training.

[0031] Step 3. By performing data augmentation on the selected dataset, including operations such as rotation, scaling, cropping, and color transformation, the diversity of samples is further enriched, and the model's ability to identify sea ice targets in different scenarios is improved.

[0032] Step 4. Use professional annotation software to accurately annotate the sea ice targets in the sea ice images, generating annotation files containing target categories and bounding box coordinates; then divide the annotated dataset into training, testing, and validation sets according to a certain ratio to meet the needs of model training and evaluation.

[0033] Step 5. By building a YOLO-based object detection algorithm framework, the model is trained using the training set, and the network parameters are continuously optimized to obtain the optimal training model suitable for sea ice identification.

[0034] Step 6. Evaluate the performance of the trained model using validation and test sets, including metrics such as detection accuracy, recall, and inference speed, to ensure the model's effectiveness and robustness in real-world applications.

[0035] Step 7. By deploying the trained and evaluated model to the actual application system, the system can automatically identify and detect newly acquired sea ice images, providing technical support for sea ice monitoring and management.

[0036] In some embodiments, the path planning algorithm in step 15 employs the Deep Q-Network (DQN) algorithm, with the following steps: S1, Initialization: Initialize the UAV system, unmanned surface vessel system, main network, target network parameters, and experience replay buffer; set hyperparameters and reward-related parameters; reset the environment and obtain the initial state (including sea ice distribution, navigation area, etc.). S2, Training Process: Loop through each training round: reset the environment, initialize the positions of the drone and unmanned surface vessel; set the total round reward and steps to zero; obtain initial sea ice distribution and navigation area information; The process iterates through each step: selecting the unmanned surface vessel's (USV) movement action according to an ε-greedy strategy; the USV performs sea ice identification and map building tasks; the USV's movement action is executed, and after each action, the environment changes, and the system receives two key pieces of information: the next state and whether the process has ended; the reward is calculated (considering factors such as proximity to the target, reaching the target, collision, and path smoothness); and the experience is stored in an experience replay buffer. The ε-greedy strategy is a classic action selection strategy in reinforcement learning (RL) used to balance the conflict between exploration and exploitation. Its core idea is: randomly selecting an action with probability ε (exploration), and selecting the currently known optimal action with probability 1-ε (exploitation). If the experience replay buffer has enough samples: extract samples; calculate the target action value function Q-value and the current action value function Q-value; calculate the loss and update the main network; The update frequency determines when the target network parameters are updated. If the update frequency is met, the target network parameters are updated. Here, meeting the update frequency means that the number of training steps in one round reaches a specified number. Update the status of drones and unmanned surface vessels, increment steps by one, and accumulate rewards; If the loop ends, record the information and exit the loop.

[0037] Throughout the reinforcement learning training cycle, each action of the unmanned surface vessel (USV) generates an instant reward. The calculation rule for this instant reward is determined by a reward function comprised of factors such as proximity to the target, reaching the target, collision, and path smoothness. All these instant rewards are accumulated in each training round to form a cumulative reward. The algorithm aims to maximize this cumulative reward by continuously adjusting the strategy, thereby enabling the USV to learn how to complete tasks efficiently and safely.

[0038] Reward function for proximity to the target The design is as follows, which makes it smoother and more effective in guiding the unmanned surface vessel towards the target: in, The current distance between the unmanned surface vessel and the target. This is the scaling factor.

[0039] Rewards are given based on the magnitude of the unmanned surface vessel's (USV) course change. Small changes result in positive rewards, while large changes result in negative rewards. The expression is as follows: in, The angle at which the unmanned surface vessel changes course. The threshold for heading change, The maximum possible angle of change in heading.

[0040] This application proposes a collaborative operation system and method for unmanned aerial vehicles (UAVs) and unmanned surface vessels (USVs), which, through a three-platform collaborative mode of UAVs, USVs, and ground stations, combined with deep learning and intelligent control technologies, achieves efficient collaborative operations in polar seas: The UAV, equipped with a depth vision camera, collects information on sea ice distribution and obstacles, constructs a digital map of the polar region, and identifies navigable areas, then transmits the data to the ground station; the ground station, based on the three-dimensional target coordinates provided by the UAV and the inertial navigation information of the USV, calculates the relative position and uses the DQN algorithm to generate an optimal path that takes into account obstacle avoidance, speed, and heading, while simultaneously sending propulsion control parameters containing speed and heading commands to the USV via wireless communication; the USV adjusts its thrust and direction according to the commands, continuously corrects its heading and speed by combining inertial navigation and real-time camera feedback, and finally accurately reaches the target position and stands by, significantly improving the automation level and mission execution accuracy of polar region monitoring.

[0041] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention. Such modifications or substitutions should all fall within the scope of the invention, or any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A collaborative operation system for unmanned aerial vehicles and surface vessels, characterized in that, The system includes: unmanned aerial vehicles, unmanned surface vessels, and ground stations; The drone includes: propellers, which provide power to the drone and are fixed to the drone's arms with screws; a drone-borne wireless communication module; and a depth vision camera, which is used to collect obstacle information in the scene. The unmanned surface vessel (USV) includes: a wireless communication module; and a propulsion system for providing power to the USV. The ground station includes: a ground station wireless communication module; and a display module for displaying the digital map transmitted by the drone. The system consists of a UAV (unmanned aerial vehicle) platform A, an unmanned surface vessel (USV) platform B, and a ground station platform C. Platform A constructs a digital map and identifies sea ice areas during operation. After completion, it transmits the data to platform C and then remains in standby mode. Platform C waits for the target position of platform B to be manually set before performing path planning and sending control commands. Platform B receives the control commands, moves to the target position according to the planned path, and then remains in standby mode. The path planning employs a Deep Q-Network (DQN) algorithm, which includes: initializing the main network, target network, and experience replay buffer; selecting actions according to an ε-greedy strategy during training; dynamically adjusting the strategy using immediate rewards from environmental feedback to maximize accumulated rewards for efficient obstacle avoidance; breaking data correlation through an experience replay mechanism; periodically updating the target network parameters; and finally outputting the planned path.

2. A method for use in the unmanned aerial vehicle (UAV) ship collaborative operation system as described in claim 1, characterized in that, The method includes: Step 1: System initialization, completing the parameter settings and function initialization of the UAV and unmanned surface vessel; Step 2: Read the initial position information according to the set initial position information; Step 3: The drone reads the target location information set by the ground station; Step 4: The drone reaches the target location using motion control algorithms; Step 5: Determine if the target location is the set target location. If the drone reaches the set target location, proceed to step 6; otherwise, proceed to step 3. Step 6: The drone reaches the set target position and hovers at a fixed altitude; Step 7: The depth vision camera carried by the drone rotates at a constant speed, and the camera captures images of objects in the current area; Step 8: Construct a digital map using the depth vision camera; Step 9: Identify sea ice in the image using a deep learning algorithm; Step 10: Mark the airworthiness area on the constructed digital map. After completion, proceed to Step 11. Step 11: Transmit map information from the drone to the ground station via wireless communication technology. After completion, proceed to step 18. Step 12: Determine whether the drone data transmission is complete. If complete, proceed to step 13; otherwise, proceed to step 11. Step 13, drone standby; Step 14: Set the destination position of the unmanned surface vessel via the ground station; Step 15: Set the unmanned surface vessel's movement path using a path planning algorithm; Step 16: Send ground station control commands to the unmanned surface vessel via wireless communication technology; Step 17: The unmanned surface vessel executes control commands and moves to the designated location; Step 18: Determine whether the unmanned surface vessel has reached the designated location. If it has, proceed to step 19; otherwise, proceed to step 16. Step 19: The system enters standby mode, waiting for new task instructions. The process ends upon completion.

3. The method according to claim 2, characterized in that, The initial position information in step 2 above includes the set flight altitude and the latitude and longitude of the specified location.

4. The method according to claim 2, characterized in that, Step 6 above includes: The drone reads the target point location set by the ground station; it then reaches the set location through motion control algorithms and hovers at a constant altitude.

5. The method according to claim 2, characterized in that, Step 9 above includes: By loading a YOLO object detection model and its weight parameters that have been pre-trained on a sea ice dataset; The system acquires images or video frames to be detected using acquisition devices to obtain raw sea ice images to be identified, and then inputs these raw sea ice images into the identification system. By performing preprocessing operations such as resizing, pixel normalization, and data format conversion on the input sea ice images, the images meet the input requirements of the YOLO object detection model, thereby improving the accuracy and efficiency of subsequent recognition. By inputting the preprocessed sea ice image into the YOLO object detection model and performing forward inference, the raw detection results containing sea ice target candidate boxes, class probabilities, and confidence scores are obtained. By post-processing the target candidate boxes output by the YOLO target detection model, including filtering valid sea ice targets based on confidence thresholds and using non-maximum suppression algorithm to remove overlapping redundant boxes, accurate sea ice detection results are finally obtained. By drawing the bounding boxes, category labels, and confidence scores of detected sea ice targets on the original sea ice image, the sea ice detection results can be visualized. The result images can be displayed on the user interface or stored in a specified location as needed, thus achieving automatic recognition of the input sea ice image.

6. The method according to claim 5, characterized in that, The YOLO object detection model is trained based on a sea ice dataset, and the training process is as follows: Step 1. Collect raw image data containing sea ice distribution using camera equipment set up in the target area, ensuring that the collection process covers diverse scenarios with different climates, lighting, and sea ice morphology changes; Step 2. Perform data cleaning on the acquired raw image data; Step 3. Perform data augmentation on the filtered dataset; Step 4. Use professional annotation software to accurately annotate the sea ice targets in the dataset, generating an annotation file containing the target category and bounding box coordinates; then divide the annotated dataset into training set, test set and validation set according to a certain ratio; Step 5. By building a target detection algorithm framework based on YOLO, the model is trained using the training set, and the network parameters are continuously optimized to obtain the optimal training model suitable for sea ice identification. Step 6. Evaluate the performance of the trained model using a validation set and a test set. The evaluation parameters include detection accuracy, recall, and inference speed. Step 7. Deploy the trained and evaluated YOLO object detection model to the actual application system for automatic recognition and detection of newly acquired sea ice images.

7. The method according to claim 6, characterized in that, Step 2 involves cleaning the acquired raw image data, including removing blurry, noisy, duplicate, or invalid image data.

8. The method according to claim 6, characterized in that, Step 3 involves performing data augmentation on the filtered dataset, including rotation, scaling, cropping, or color transformation.

9. The method according to claim 5, characterized in that, The path planning algorithm uses the Deep Q-Network (DQN) algorithm, and the steps are as follows: S1, Initialization: Initialize the UAV system, unmanned surface vessel system, main network, target network parameters, and experience replay buffer; set hyperparameters and reward-related parameters; reset the environment and obtain the initial state; S2, Training Process: Loop through each training round: reset the environment, initialize the positions of the drone and unmanned surface vessel; set the total round reward and steps to zero; obtain initial sea ice distribution and navigation area information; The process iterates through each step: selecting the unmanned surface vessel's movement action according to an ε-greedy strategy; the unmanned surface vessel performs sea ice identification and map building tasks; the unmanned surface vessel's movement action is executed, and two key pieces of information are obtained: the next state and whether the process has ended; the reward is calculated; and the experience is stored in the experience replay buffer. If the experience replay buffer has enough samples: extract samples; Calculate the target action value function Q-value and the current action value function Q-value; Calculate the loss and update the main network; If the update frequency is met, the target network parameters are updated. The update frequency means that the number of training steps in one round reaches a specified number. Update the status of drones and unmanned surface vessels, increment steps by one, and accumulate rewards; If the loop ends, record the information and exit the loop. In the entire reinforcement learning training loop, each action of the unmanned surface vessel (USV) generates an immediate reward, which is determined by a reward function. The immediate rewards are accumulated in each training round to form a cumulative reward. The goal of the algorithm is to maximize the cumulative reward by continuously adjusting the strategy, thereby enabling the USV to learn to complete tasks efficiently and safely. Reward function for proximity to the target The design is as follows: In the formula, The current distance between the unmanned surface vessel and the target. This is the scaling factor; Rewards are also given based on the magnitude of the unmanned surface vessel's (USV) course change. Small course changes result in positive rewards, while larger changes result in negative rewards. The reward formula is as follows: In the formula, The angle at which the unmanned surface vessel changes course. The threshold for heading change, This represents the angle of maximum heading change.

Images