Real-time positioning, tracking and precise launching system for sea rescue unmanned aerial vehicle

Abstract

The invention relates to the technical field of sea rescue, in particular to a real-time positioning, tracking and precise launching system for a sea rescue unmanned aerial vehicle. Comprising a multi-sensor fusion positioning module, and a GPS positioning module, an IMU and a visual sensor are adopted to collect data and perform fusion positioning; the real-time target tracking module is used for identifying and tracking a maritime distress target in real time by using a visual sensor and automatically tracking the target in combination with an unmanned aerial vehicle flight control system; according to the material precise throwing module, an adjustable throwing device is installed on the unmanned aerial vehicle, the angle and force of the throwing device are adjusted through precise calculation and control, and precise throwing of materials is achieved; and the data communication and remote control module is used for establishing a high-speed data communication link between the unmanned aerial vehicle and a ground control center and realizing real-time transmission and remote control of data. According to the invention, the positioning precision, the target tracking real-time performance and the material putting precision can be improved, and the sea rescue efficiency and success rate can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of marine rescue, and in particular to a real-time positioning tracking and accurate delivery system for a marine rescue unmanned aerial vehicle. BACKGROUND

[0002] Marine rescue work has always been a challenging task. In the vast ocean environment, after a maritime accident occurs, rescue work faces multiple difficulties such as complex environment and dynamic changes of targets, and the current lack of efficient and accurate search and rescue means directly affects the rescue efficiency and survival probability of the people in distress.

[0003] From the existing rescue means, the traditional rescue ship is limited by the sailing speed, and when facing the vast ocean, especially the high sea area far from the coastline, the response lag problem is prominent. For example, small boats in remote waters in distress, rescue ships often need several hours or even several days to arrive, missing the best rescue opportunity; although the helicopter for air search and rescue has speed advantage, it is greatly affected by air flow and weather, and the delivery of materials is easy to deviate from the predetermined position, and its flight cost is high and the endurance is limited, which cannot perform the task for a long time; the traditional satellite positioning can only provide a rough position, and the precision is insufficient for small targets floating on the sea or blocked targets, which cannot match the positioning needs of the people in distress moving dynamically with the sea waves and sea currents, resulting in blind delivery of rescue materials.

[0004] The timeliness of marine rescue is directly related to the life safety of people in distress, who face survival threats such as cold, hunger, and dehydration on the sea. Every minute of delay increases the risk of life. Therefore, how to break through the problems such as slow response of rescue ships, insufficient delivery precision of air search and rescue, and limitations of traditional positioning means, and effectively improve the efficiency and success rate of marine rescue is a problem to be solved. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a real-time positioning tracking and accurate delivery system for a marine rescue unmanned aerial vehicle to solve the problems of slow response, large delivery deviation, and insufficient positioning of the existing rescue means, which affect the efficiency and success rate of marine rescue.

[0006] In order to achieve the above purpose, the present application provides a real-time positioning tracking and accurate delivery system for a marine rescue unmanned aerial vehicle, which comprises:

[0007] A multi-sensor fusion positioning module comprising a GPS positioning module, an IMU, and a vision sensor, the GPS positioning module, the IMU, and the vision sensor collect data, and then fuse the data through a data fusion algorithm, and when the GPS positioning module is disturbed, the data of the IMU and the vision sensor are combined to optimize the positioning to obtain the position and attitude information of the unmanned aerial vehicle;

[0008] A real-time target tracking module, based on the position and attitude information, uses a visual sensor to identify and track the maritime distress target in real time, and realizes automatic tracking of the target by combining the unmanned aerial vehicle flight control system to obtain target dynamic information;

[0009] A material precise delivery module, including a delivery device installed on the unmanned aerial vehicle, based on the target dynamic information, according to the relative position, distance, wind speed and flight speed of the unmanned aerial vehicle and the target, a delivery parameter calculation model is constructed based on the kinematics principle to solve the material falling time and the required initial horizontal velocity component; then the parameters are solved through the correlation equation set of the initial velocity, the delivery angle and the delivery strength, the pitch angle, the yaw angle and the delivery strength of the delivery device are adjusted and real-time corrected;

[0010] A data communication and remote control module, using multi-band hybrid communication technology and combining diversity technology to establish a high-speed data communication link, transmitting positioning, tracking and delivery related data through data compression and error correction coding, and using encryption, priority scheduling and real-time monitoring retransmission mechanism for remote control instructions.

[0011] Preferably, the multi-sensor fusion positioning module needs to calibrate the GPS positioning module, IMU and visual sensor before being enabled, and the calibration process includes static calibration and dynamic calibration. The static calibration is to adjust the zero point and sensitivity of the sensor, and the dynamic calibration is to real-time correct the sensor data in the actual flight process.

[0012] Preferably, the GPS positioning module uses satellite signals to determine the absolute position of the unmanned aerial vehicle; the IMU is composed of an accelerometer and a gyroscope, the accelerometer is used to measure the acceleration of the unmanned aerial vehicle, and the gyroscope is used to measure the angular velocity of the unmanned aerial vehicle; the visual sensor uses optical imaging principle to obtain image information of the maritime environment, and identifies the target object on the sea through image processing algorithm.

[0013] Preferably, the data fusion algorithm is Kalman filter algorithm, which can make optimal estimation of the state of the system based on the state equation and the observation equation , wherein represents time, represents the state vector of the unmanned aerial vehicle at the time, is the state transition matrix of the unmanned aerial vehicle at the time, is the control input matrix of the unmanned aerial vehicle at the time, is the control input vector of the unmanned aerial vehicle at the time, is the process noise of the unmanned aerial vehicle at the time, and . a process noise covariance matrix of the UAV at the kth moment, a process noise covariance matrix of the UAV at the kth moment, an observation vector of the UAV at the kth moment, an observation vector of the UAV at the kth moment, an observation matrix of the UAV at the kth moment, an observation matrix of the UAV at the kth moment, an observation noise of the UAV at the kth moment, and an observation noise of the UAV at the kth moment, and an observation noise covariance matrix of the UAV at the kth moment. an observation noise covariance matrix of the UAV at the kth moment.

[0014] Preferably, the specific operation of optimizing positioning in combination with IMU and visual sensor data when the GPS positioning module is disturbed is that when the signal of the GPS positioning module is disturbed, the weight of the IMU and visual sensor data in fusion is increased, and the result of the dead reckoning is fused and optimized with the result of the Kalman filtering by using the data of the IMU and visual sensor.

[0015] Preferably, the real-time identification and tracking of the marine distress target by the visual sensor adopts an improved YOLOv5 deep learning target identification algorithm based on CNN and an optimized tracking method based on KCF algorithm and combined with multi-scale search and adaptive learning rate adjustment.

[0016] Preferably, the multi-band hybrid communication technology dynamically adjusts the communication proportion of the microwave and millimeter wave frequency bands through the weight coefficient The communication proportion of the microwave and millimeter wave frequency bands is dynamically adjusted, the millimeter wave weight is increased for close-range communication, and the microwave weight is increased for long-distance or shielding scenes.

[0017] The present application has the following advantages: compared with the prior art, the present application improves the positioning accuracy, target tracking real-time performance and material dropping accuracy through multi-module cooperation, and effectively improves the efficiency and success rate of marine rescue. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only illustrate the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0019] Figure 1 The present application is a flowchart. DETAILED DESCRIPTION

[0020] ​For purposes of the present invention, the technical solutions and beneficial effects are more clearly apparent, the following detailed description of the present invention with specific examples. It should be noted that, unless otherwise defined, the present invention using technical terms or scientific terms should be understood by those skilled in the art of the present invention belongs to the general meaning of the people.

[0021] As Figure 1 shown, a real-time positioning and tracking and accurate delivery system for offshore rescue unmanned aerial vehicle, comprising:

[0022] 1, multi-sensor fusion positioning module

[0023] 1.1, sensor selection and principle

[0024] In the real-time positioning and tracking and accurate delivery system for offshore rescue unmanned aerial vehicle, a high-precision GPS (Global Positioning System, Global Positioning System) positioning module, IMU (Inertial Measurement Unit, Inertial Measurement Unit) and vision sensor are used for fusion positioning. GPS positioning module uses satellite signals to determine the absolute position of the unmanned aerial vehicle, its basic principle is to measure the propagation time of satellite signals to calculate the distance between the unmanned aerial vehicle and the satellite, and then determine the position of the unmanned aerial vehicle on the earth. Set the propagation time of the first satellite signal, the speed of light, the distance between the unmanned aerial vehicle and the first satellite, then . Through the distance measurement of at least four satellites, the equation group can be established to solve the three-dimensional position of the unmanned aerial vehicle .

[0025] IMU mainly consists of an accelerometer and a gyroscope. The accelerometer is used to measure the acceleration of the unmanned aerial vehicle, and the gyroscope is used to measure the angular velocity of the unmanned aerial vehicle. Set are the measurement values of the accelerometer on the axis, are the measurement values of the gyroscope on the axis. By integrating the accelerometer measurement values, the velocity information can be obtained, and by further integrating, the position information can be obtained; by integrating the gyroscope measurement values, the attitude information can be obtained.

[0026] The vision sensor uses optical imaging principle to obtain image information of the sea environment, and identifies the target object on the sea through image processing algorithm. The vision sensor has the advantage of providing rich scene information, which can be used to supplement the deficiencies of GPS positioning module and IMU in some cases.

[0027] 1.2, data fusion algorithm

[0028] To improve positioning accuracy and reliability, data from the GPS positioning module, IMU, and visual sensor need to be fused. Here, the Kalman filter algorithm is used for data fusion. The Kalman filter is an optimal recursive filter that can optimally estimate the system state based on the system's state equation and observation equation.

[0029] Let the state vector of the UAV be... ,in Indicates time, Indicates the first The state vector of the drone at any given time. For the first The position vector of the drone at any given time. For the first The velocity vector of the drone at any given moment. For the first The attitude quaternion of the drone at any given time. The state equation of the system can be expressed as:

[0030]

[0031] in, For the first The state transition matrix of the drone at any time. For the first The control input matrix of the drone at any time For the first The control input vector of the drone at any time. For the first The process noise of the drone at all times, and , For the first The process noise covariance matrix of the drone at any given time.

[0032] The observation equation can be expressed as:

[0033]

[0034] in, For the first The observation vector of the drone at any time. For the first The observation matrix of the drone at any time For the first The observation noise of the drone at all times, and , For the first The observation noise covariance matrix of the drone.

[0035] The Kalman filtering process includes prediction and updating. The prediction step is as follows:

[0036]

[0037]

[0038] wherein, is the prior state estimate of the UAV at time is the posterior state estimate of the UAV at time is the prior error covariance matrix of the UAV at time is the posterior error covariance matrix of the UAV at time is the transpose matrix of

[0039] The update step is:

[0040]

[0041]

[0042]

[0043] wherein, is the Kalman gain of the UAV at time is the updated state estimate of the UAV at time is the updated error covariance matrix of the UAV at time 1.3、Strategy for dealing with GPS signal interference

[0044] In the marine environment, the GPS signal can be interfered, resulting in inaccurate positioning. In order to deal with this situation, when the GPS signal is interfered, the weight of the IMU and visual sensor data in fusion can be increased. Let

[0045] be the weight of the GPS positioning module, the IMU and the visual sensor data in fusion respectively, and When the GPS signal quality is detected to be degraded, the is reduced, and the , the and the are increased.

[0046] ​​​​​​​Meanwhile, dead reckoning is performed using data from the IMU and visual sensor. According to the acceleration and angular velocity measured by the IMU, combined with the initial position and attitude information, the position and attitude changes of the UAV are calculated by integration. The visual sensor identifies feature points on the sea and estimates the motion of the UAV using visual odometry technology. The results of dead reckoning are fused with the results of Kalman filtering to further improve the accuracy of positioning.

[0047] Before use, the GPS positioning module, IMU and visual sensor are calibrated to ensure the accuracy of sensor data. The calibration process includes static calibration and dynamic calibration. Static calibration mainly adjusts the zero point and sensitivity of the sensor, and dynamic calibration corrects the sensor data in real time during actual flight. Through the above multi-sensor fusion positioning module, the positioning accuracy and reliability of the UAV in complex marine environments can be improved, providing a solid foundation for subsequent target tracking and precise delivery of supplies.

[0048] 2. Real-time target tracking module

[0049] 2.1. Visual sensor target recognition algorithm

[0050] In the marine rescue scenario, the visual sensor needs to accurately identify the distress target in a complex marine background. An improved YOLOv5 deep learning target recognition algorithm based on Convolutional Neural Network (CNN) is used. This algorithm can detect targets at different resolutions through multi-scale feature fusion, improving the accuracy of target recognition.

[0051] Let the input image be where is the image height, is the image width, is the number of channels. After the CNN feature extraction process, a series of feature maps are obtained, each feature map , , , is the height, width and number of channels of the th feature map, respectively.

[0052] In YOLOv5, an anchor box mechanism is used to predict the position and class of the target. For each grid on each feature map, a total of anchor boxes of different sizes and proportions are predefined. Let the prediction output of the th anchor box on the th grid of the th feature map be , where , where, are the log-transformed values of the width and height of the prediction box, is the target confidence, is the probability of the class.

[0053] The actual position and size of the prediction box can be calculated by the following equations:

[0054]

[0055]

[0056]

[0057]

[0058] where, are the center coordinates of the prediction box, are the width and height of the prediction box, are the top-left corner coordinates of the grid, are the initial width and height of the anchor box, is the Sigmoid function.

[0059] 2.2. Target tracking model

[0060] After identifying the target, a target tracking algorithm based on correlation filtering is used for real-time tracking. The Kernel Correlation Filter (KCF) algorithm is used as the basis, combined with multi-scale search and adaptive learning rate adjustment.

[0061] Let the target template be and the search region be . The KCF algorithm calculates the correlation matrix of the target template and the search region to obtain the correlation filter response. The correlation filter response can be represented as:

[0062]

[0063] where, are the Fourier transform and inverse Fourier transform respectively, is element-wise multiplication,

[0064] ​​​​​​​​​​​​​​​​​​​​​​​​​​​To adapt to the scale changes of the target, a multi-scale search method is adopted. Let the initial scale be , the scale factor be , and the scale set for search be . For each scale , the search area is scaled and the correlation filter response is calculated. The scale with the maximum response value is selected as the scale of the current target.

[0065] To improve the stability of tracking, an adaptive learning rate adjustment is adopted. Let the learning rate be , and dynamically adjust the learning rate according to the motion state of the target and the confidence of tracking during the tracking process. When the target moves faster or the tracking confidence is lower, increase the learning rate to update the target template faster; when the target moves slower or the tracking confidence is higher, reduce the learning rate to maintain the stability of the target template.

[0066] 2.3, Target tracking combined with flight control system

[0067] The target tracking result of the visual sensor is combined with the flight control system of the unmanned aerial vehicle to realize automatic tracking of the target. Let the position of the target be , the current position of the unmanned aerial vehicle be , and the flight speed of the unmanned aerial vehicle be .

[0068] According to the positional relationship between the target and the unmanned aerial vehicle, the flight direction and speed adjustment of the unmanned aerial vehicle are calculated. The flight direction can be calculated by the following formula:

[0069]

[0070] The speed adjustment of the unmanned aerial vehicle is adjusted according to the distance between the target and the unmanned aerial vehicle. When the distance is large, the flight speed is increased; when the distance is small, the flight speed is reduced to ensure that the unmanned aerial vehicle can stably track the target.

[0071] Through the above details, the real-time target tracking module can accurately identify and track the distress target in complex marine environment, and combine the flight control system of the unmanned aerial vehicle to realize automatic tracking of the target, providing reliable target dynamic information for subsequent precise material delivery.

[0072] 3, Precise material delivery module

[0073] In the marine rescue scenario, the precise delivery module of supplies is crucial to improve the efficiency and success rate of rescue. First, install an adjustable delivery device on the UAV, then consider various factors such as the relative position and distance of the UAV and the target, wind speed, etc. Through precise calculation and control, adjust the angle and strength of the delivery device, and finally ensure that the rescue supplies can accurately fall near the target location.

[0074] 3.1 Delivery parameter calculation model

[0075] Let the position of the UAV in three-dimensional space be , and the target position be , then the horizontal distance and the vertical distance between the UAV and the target are:

[0076]

[0077]

[0078] Let the wind speed vector be , and the UAV flight speed vector be . After the delivery of supplies, in the horizontal direction, the supplies are affected by the wind speed and their initial horizontal speed; in the vertical direction, the supplies are affected by gravity.

[0079] According to the kinematics principle, the motion equation of the supplies in the vertical direction is:

[0080]

[0081] where is the vertical position of the supplies at time , and is the acceleration of gravity. When the supplies reach the target vertical height , i.e. , the falling time of the supplies can be solved:

[0082]

[0083] In the horizontal direction, the horizontal displacement and of the supplies are:

[0084]

[0085]

[0086] In order to make the supplies accurately reach the target position , it is necessary to adjust the initial horizontal speed component and such that:

[0087]

[0088]

[0089] Solving the above equation set can obtain:

[0090]

[0091]

[0092] 3.2, the angle and strength adjustment of the delivery device

[0093] The angle and strength of the delivery device directly affect the initial speed of the supplies. Let the pitch angle of the delivery device be , the yaw angle be , and the delivery strength be .

[0094] The initial speed vector of the supplies is The relationship between the initial speed vector and the delivery angle and strength is:

[0095]

[0096]

[0097]

[0098] Combining the and calculated above, an equation set can be established to solve the delivery angle and and the delivery strength :

[0099]

[0100] By solving the above equation set, the appropriate delivery angle , and strength can be obtained. The unmanned aerial vehicle controls the delivery device according to the calculated parameters, and the rescue supplies are accurately delivered to the target, so as to realize the accurate delivery of the supplies.

[0101] 3.3, real-time correction mechanism

[0102] During the delivery of the supplies, the wind speed, the flight state of the unmanned aerial vehicle, and other factors may change, and a real-time correction mechanism needs to be established.

[0103] The ground control center receives real-time location information from drones and supplies, and compares the actual location of the supplies with the expected location. , and The parameters of the delivery device are adjusted in real time. Let the corrected delivery angle be... yaw angle is The investment intensity is Then, a correction formula can be established based on the deviation information:

[0104]

[0105]

[0106]

[0107] in, The correction factor can be adjusted according to the actual situation. A real-time correction mechanism can further improve the accuracy of resource allocation.

[0108] The precise delivery module for supplies achieves accurate delivery by establishing a precise delivery parameter calculation model, adjusting the angle and force of the delivery device, and introducing a real-time correction mechanism. It fully considers the complex factors of maritime rescue scenarios and provides strong methodological support for maritime rescue work.

[0109] 4. Data communication and remote control module

[0110] 4.1 Establishment of high-speed data communication links

[0111] In maritime rescue drone systems, multi-band hybrid communication technology is employed to achieve high-speed, stable, and reliable data communication between the drone and the ground control center. This method combines the advantages of microwave and millimeter-wave bands to adapt to different communication distances and environmental requirements.

[0112] Assume the communication rate in the microwave band is The communication rate in the millimeter wave band is Total communication rate of hybrid communication links It can be represented as:

[0113]

[0114] in, This represents the weight of microwave frequency band communication rate in the total communication rate. In short-range communication scenarios, the millimeter-wave band offers higher bandwidth and data transmission rates, allowing for appropriate increases in bandwidth. The value; however, in long-distance communication or under obstructed conditions, the microwave band offers better communication stability and can increase the... The value of the data is determined by the value of the data.

[0115] At the same time, in order to improve the reliability of communication, diversity technology is adopted. By setting multiple antennas on the unmanned aerial vehicle and the ground control center respectively, the influence of signal fading is reduced by using the principle of space diversity and frequency diversity. Let the signal strength received by the receiving end of the first antenna be , and the signal strength after diversity combination be The signal strength after diversity combination can be calculated by the maximum ratio combination algorithm:

[0116]

[0117] wherein, is the number of antennas, is the weighting coefficient of the first antenna, which is proportional to the signal-to-noise ratio of the signal received by the antenna.

[0118] 4.2 Real-time data transmission

[0119] The various data collected by the unmanned aerial vehicle, such as position information , attitude information (pitch angle , roll angle , yaw angle ), sensor data (such as image data of visual sensor, acceleration and angular velocity data of IMU) and related information of the target (such as position, speed, etc.), need to be transmitted to the ground control center in real time and accurately.

[0120] In order to improve the efficiency and reliability of data transmission, data compression and error correction coding technology is adopted. For image data, advanced image compression algorithms such as JPEG2000 are adopted to compress the original image data to times, of the original. For other data, lossless compression algorithms such as Huffman coding are used for compression.

[0121] During transmission, error correction coding technology such as cyclic redundancy check (CRC) is adopted to check and correct the transmitted data. Let the original data be , the data after CRC encoding be , and the encoding process can be represented as:

[0122]

[0123] wherein, is the generating polynomial, is the remainder. After receiving the data, the receiving end compares the remainder with the received remainder to determine whether the data is transmitted incorrectly and makes the corresponding correction.

[0124] 4.3 Remote Control

[0125] The ground control center remotely controls the UAV based on the received information about the UAV and the target. The remote control instructions include flight trajectory adjustment instructions (such as target point coordinates ), drop parameter adjustment instructions (such as drop angle , drop intensity ), etc.

[0126] When transmitting remote control instructions, visual odometry technology is used to ensure the safety and reliability of the instructions. The control instructions are encrypted to generate encrypted instructions :

[0127]

[0128] where, is the original control instruction, is the encryption key, is the encryption algorithm. After receiving the encrypted instructions, the UAV uses the corresponding decryption algorithm to decrypt:

[0129]

[0130] At the same time, in order to ensure the real-time nature of the control instructions, a priority scheduling algorithm is used. According to the urgency and importance of the instructions, different priorities are assigned to different control instructions. Let the priority of instruction be , the scheduling algorithm will prioritize instructions with high priority to ensure that the flight state and drop parameters of the UAV can be adjusted in time in emergency situations.

[0131] In complex marine environments, due to factors such as signal interference and delay, the loss or error of control instructions may occur. Therefore, the ground control center will monitor and retransmit the instructions in real time. Let the sending times of instruction be , when no confirmation information is received from the UAV within a specified time, the instruction will be re-sent until the confirmation information is received or the maximum sending times are reached.

[0132] Those skilled in the art should understand that the above discussion of any embodiment is only intended to be illustrative and is not intended to be in any way limiting as to the scope of the present application, including the claims that follow it; the above embodiments or technical features among different embodiments can also be combined, steps can be implemented in any order, and there are many other changes to the different aspects of the present application as described above, which are not provided in details for the sake of brevity.

[0133] The present application is intended to cover all such alternatives, modifications, and variations as fall within the broad scope of the appended claims. Accordingly, any one of the above-described embodiments of the present application can be further modified than or combined with another in addition to the changes discussed in the above description.

Claims

1. A real-time positioning tracking and precise dropping system for offshore rescue drones, characterized in that, The application relates to a multi-sensor fusion positioning module, a real-time target tracking module and a material accurate delivery module. The multi-sensor fusion positioning module comprises a GPS positioning module, an IMU and a visual sensor, the GPS positioning module, the IMU and the visual sensor collect data, and then the data are fused through a data fusion algorithm; when the GPS positioning module is interfered, the data of the IMU and the visual sensor are combined to optimize positioning, so that the position and attitude information of the unmanned aerial vehicle is obtained. The real-time target tracking module is based on the position and attitude information, uses the visual sensor to realize real-time identification and tracking of a marine distress target, and realizes automatic tracking of the target by combining an unmanned aerial vehicle flight control system to obtain target dynamic information. The material accurate delivery module comprises a delivery device installed on the unmanned aerial vehicle, and is based on the target dynamic information, the relative position and distance between the unmanned aerial vehicle and the target, wind speed and unmanned aerial vehicle flight speed, a delivery parameter calculation model is constructed based on kinematics principle to solve the falling time of the material and the required initial horizontal velocity component; then the parameters are solved through a correlation equation set of the initial velocity, the delivery angle and the delivery strength, the pitch angle, the yaw angle and the delivery strength of the delivery device are adjusted and real-time correction is carried out. The data communication and remote control module adopts a multi-band mixed communication technology and combines a diversity technology to establish a high-speed data communication link, transmits positioning, tracking and delivery related data through data compression and error correction coding, and adopts encryption, priority scheduling and real-time monitoring and retransmission mechanism for remote control instructions.

2. The system of claim 1, wherein, The multi-sensor fusion positioning module needs to calibrate the GPS positioning module, the IMU and the visual sensor before being started, the calibration process comprises static calibration and dynamic calibration, the static calibration is to adjust the zero point and sensitivity of the sensor, and the dynamic calibration is to real-time correct the data of the sensor in the actual flight process.

3. The system of claim 1 or 2, wherein, The GPS positioning module uses satellite signals to determine the absolute position of the unmanned aerial vehicle; the IMU is composed of an accelerometer and a gyroscope, the accelerometer is used for measuring the acceleration of the unmanned aerial vehicle, and the gyroscope is used for measuring the angular velocity of the unmanned aerial vehicle. The visual sensor uses optical imaging principle to obtain image information of the marine environment, and identifies the target object on the sea through an image processing algorithm.

4. The system of claim 1, wherein, The data fusion algorithm is a Kalman filter algorithm, which can be based on the system's state equations. and observation equations To perform an optimal estimate of the system's state, where Indicates time, Indicates the first The state vector of the drone at any given time. For the first The state transition matrix of the drone at any time. For the first The control input matrix of the drone at any time For the first The control input vector of the drone at any time. For the first The process noise of the drone at all times, and , For the first The process noise covariance matrix of the drone at any given time. For the first The observation vector of the drone at any time. For the first The observation matrix of the drone at any time For the first The observation noise of the drone at all times, and , For the first The observation noise covariance matrix of the drone.

5. The system of claim 1, wherein, When the GPS positioning module is interfered, the weight of the data of the IMU and the visual sensor in the fusion is increased, and the data of the IMU and the visual sensor are used for track prediction, and the track prediction result is fused with the result of Kalman filtering to optimize positioning.

6. The system of claim 1, wherein, The visual sensor uses an improved YOLOv5 deep learning target identification algorithm based on CNN and an optimized tracking method based on KCF algorithm and combined with multi-scale search and adaptive learning rate adjustment.

7. The system of claim 1, wherein, The multi-band hybrid communication technology adjusts the communication proportion of the microwave and millimeter wave frequency bands by a weight coefficient The communication proportion of the microwave and millimeter wave frequency bands is dynamically adjusted, the millimeter wave weight is increased in close-range communication, and the microwave weight is increased in long-distance or shielding scenarios.

Images