A joint design method and system for water flow estimation and ship waterline monitoring

Abstract

The present invention provides a jointly designed method and system for current estimation and ship waterline monitoring, belonging to the field of ship monitoring technology. The method comprises the following steps: First, a shore-based center deploys an underwater robot (AUV) based on the monitoring mission requirements, plans its motion path, and establishes a water gauge dataset for model training. Upon reaching the vicinity of the ship, the AUV uses a Doppler velocimeter to measure the water velocity and direction. Combined with the AUV's trajectory information, the AUV predicts the future water velocity field. Simultaneously, the AUV captures waterline images using an onboard camera, performs preprocessing, and calculates the waterline image's reliability. Furthermore, the AUV corrects its posture in real time based on the predicted water velocity field and the waterline image's reliability, achieving joint estimation of the current and the ship's waterline. This invention combines current estimation with ship waterline monitoring, enabling stable, efficient, and comprehensive monitoring of the ship's waterline using the AUV.

Description

Technical Field

[0001] The present invention belongs to the technical field of ship monitoring, and particularly relates to a joint design method and system for water flow estimation and ship waterline monitoring. Background Art

[0002] Ship safety is becoming a growing concern within the shipping industry. Waterline monitoring, a key indicator of a ship's floating stability and structural safety, currently relies primarily on manual visual inspection. While this method has proven effective in ensuring ship safety, it is costly, inefficient, and incapable of providing all-weather, all-encompassing, real-time waterline monitoring. To meet the growing demand for intelligent ports and promote their development, the use of underwater robots for ship waterline monitoring is urgently needed.

[0003] In the prior art, the publication number is CN118366016A, entitled "A method for intelligent monitoring of ships based on cross-domain perception of underwater robots", which discloses a method for intelligent monitoring of ships based on cross-domain perception of underwater robots. This method first sets the ship's markers and establishes a marker data set. Then, the underwater robot is deployed to the monitoring waters to perform the marker monitoring task. The onboard camera is used to capture the hull marker image, and then the captured surface marker image is segmented to extract the feature information of the image, and the marker is detected through the model. However, this method does not take into account the impact of water flow on the underwater robot. The water flow will cause the underwater robot's posture to change in real time, reducing reliability and making it impossible to monitor the waterline in real time.

[0004] Furthermore, the publication number CN110749890A, entitled "A Method for Collaborative Current Estimation by Multiple Underwater Robots," discloses a method for collaborative current estimation by multiple underwater robots. The method uses the position and distance information of the surface mother ship and the position estimation information, current estimation information, and distance information of the other underwater robots; it integrates the current estimation results with the Doppler velocity measurement information to compensate for the position error caused by the Doppler velocity measurement, thereby achieving underwater robot position estimation. This method estimates water flow based on the status information and velocity measurement information between multiple underwater robots. However, it does not jointly design water flow estimation with the actual monitoring task, resulting in competition for communication and computing resources. In order to achieve stable and efficient monitoring of the ship's waterline by underwater robots, it is urgently necessary to design a joint design method for water flow estimation and ship waterline monitoring. Summary of the Invention

[0005] To address the low reliability of existing technologies, which prevent real-time monitoring of a vessel's waterline, and the lack of integrated design of current estimation with the actual monitoring task, leading to competition for communication and computing resources, the present invention provides a method and system for integrated design of current estimation and vessel waterline monitoring, achieving the goal of stable and efficient monitoring of a vessel's waterline by an underwater robot.

[0006] The technical solution of the combined design method and system for water flow estimation and ship waterline monitoring of the present invention is as follows:

[0007] A joint design method for water flow estimation and ship waterline monitoring is characterized by comprising the following steps:

[0008] S1. Initialize the underwater robot and select the ship to be monitored. Match the number of underwater robots to the number of water gauges to be monitored. The shore-based control center divides the monitoring area for each underwater robot according to the monitoring task requirements. Specifically, set the monitoring target as the ship's waterline. The underwater robot collects images of the ship's waterline and establishes a water gauge dataset. The water gauge dataset is trained using a neural network to obtain a deep learning model and the underwater robot is deployed to the mission waters. The underwater robot performs the monitoring task according to the path and monitoring area planned by the shore-based control center.

[0009] S2. After the underwater robot enters the monitoring area, it measures the water flow using a Doppler velocimeter, collecting both velocity and direction data at predetermined intervals. The actual trajectory of the underwater robot is compared with the path planned by the shore-based control center, and a deviation is established. The position, velocity, and attitude of the underwater robot are then obtained using an inertial navigation and positioning system.

[0010] S3. Each underwater robot communicates via broadcasting, sending and receiving water flow velocity data from other local areas, and integrates the current trajectory deviation and flow velocity data through Kalman filtering to predict the future flow velocity field. The underwater robot also collects waterline images in real time, preprocesses the collected waterline images, and then preprocesses the waterline images to obtain credibility.

[0011] S4. The underwater robot collects its own position, velocity, and attitude data through the inertial navigation positioning device to obtain its current attitude, and uses the waterline image credibility and the water flow velocity field prediction as feedback signals to correct its attitude. When the waterline image credibility reaches a threshold, the underwater robot adjusts its attitude through the water flow velocity field prediction to minimize the attitude optimization function and achieve stable monitoring. If the waterline image credibility does not reach the threshold, the underwater robot continues to search and adjust its attitude along the path planned by the shore-based control center until the waterline image credibility reaches the threshold.

[0012] S5. Using an edge detection algorithm to identify edges in the preprocessed waterline image, and employing double threshold segmentation to separate the waterline in the image; specifically, the underwater robot uses a deep learning algorithm to identify the draft gauge number, combines it with the separated waterline, and obtains the position of the waterline on the draft gauge. The draft value of the ship is then obtained using a draft gauge reading algorithm and uploaded to the shore-based control center.

[0013] S6. According to the monitoring task requirements, the underwater robot (4) determines whether the monitoring task is completed. If not, it continues to perform monitoring. If the task is completed, the shore-based control center (1) issues a command to control the underwater robot (4) to return.

[0014] A further improvement of the technical solution of the present invention is that: in step S1, a water gauge dataset is established for training the deep learning model. The monitored ship water gauges are water gauges drawn on both sides of a large ship, distributed at the bow, midship, and stern, as specifically expressed by the following formula:

[0015]

[0016] Among them, C is the ship's water gauge reading, H d is the height of the water gauge character spacing, F is the character reading closest to the waterline, H w It is the height from the waterline to the nearest character.

[0017] A further improvement of the technical solution of the present invention is that: in step S2, the water flow is measured by a Doppler velocimeter, and the water flow velocity and flow direction data are collected simultaneously within a predetermined time interval, specifically,

[0018] At preset time intervals, the Doppler velocimeter emits sound wave pulses and receives sound waves reflected from the seabed or particles in the water. The speed and direction of the water flow are determined by calculating the changes in the sound wave frequency. The formula is:

[0019]

[0020] Where v is the water velocity, Δf is the Doppler frequency shift, c is the speed of sound waves in water, f0 is the frequency of the sound waves emitted by the Doppler velocimeter, and θ is the angle between the direction of sound wave emission and the direction of water flow.

[0021] A further improvement of the technical solution of the present invention is that in step S2, the actual running trajectory of the underwater robot is compared with the path planned by the shore-based control center to establish a deviation, specifically,

[0022] After collecting the water flow velocity and direction data, the underwater robot will compare this data with its actual motion trajectory and establish the trajectory deviation Δd:

[0023]

[0024] Where Δd is the trajectory deviation, (x a ,x p ) is the actual position coordinate of the underwater robot, (y a ,y p ) are the predetermined position coordinates of the underwater robot.

[0025] A further improvement of the technical solution of the present invention is that in step S3, each underwater robot communicates by broadcasting, sends and receives other local water flow velocity data, and integrates the current trajectory deviation and flow velocity data through Kalman filtering to predict the future flow velocity field, specifically,

[0026] Assume that the flow velocity data collected by each underwater robot is v i (t), where i represents the i-th underwater robot and t represents time. Set the initial state estimate And covariance matrix P0, predict the state and covariance of the next moment according to the underwater robot dynamics model:

[0027]

[0028] Where k represents the time step, is the predicted state matrix at time step k, F k is the state transition matrix, is the optimal estimated state matrix at time step k-1, B k is the control matrix, u k is the control input matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k, P k-1 is the covariance at time step k-1, is the transpose of the state transfer matrix, Q k is the process noise covariance matrix;

[0029] When there is new measurement data z k When , update the state estimate:

[0030]

[0031] P k =(IK k H k )P k|k-1

[0032] Among them, K k is the Kalman gain matrix, P k|k-1 is the prediction covariance matrix at time step k, H k is the observation matrix, is the transpose of the observation matrix, R k is the observation noise covariance matrix, is the updated state estimate matrix at time step k, is the predicted state matrix at time step k, I is the identity matrix, and P k The updated covariance matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k.

[0033] A further improvement of the technical solution of the present invention is that: in step S3, the collected waterline image is preprocessed, and then the credibility of the waterline image is calculated based on the preprocessed image, including the following steps:

[0034] S3.1. Image denoising: Images often contain noise. To reduce random noise and improve image clarity, we use the Gaussian kernel function to convolve the image and remove the Gaussian noise.

[0035] S3.2. Enhance contrast, improve the visual effect of the image, and make the image details clearer. Use histogram equalization as follows:

[0036]

[0037] Where r(k) is the grayscale of the input image, T(r(k)) is the cumulative probability of grayscale r(k), s(k) is the grayscale of the output image, k is the grayscale of the pixel in the image, and n j is the pixel value of gray level j, n is the total pixel value;

[0038] S3.3. After image preprocessing, the waterline image credibility C is calculated:

[0039]

[0040] Among them, N e is the number of edge pixels detected, N t is the total number of pixels in the image, is the variance of the image noise.

[0041] A further improvement of the technical solution of the present invention is that in step S4, the underwater robot collects position, speed and posture data, as well as the waterline image credibility, and performs joint posture correction to make the waterline image credibility reach a threshold and achieve posture stability. The optimization function is expressed as follows:

[0042]

[0043] in, is the minimization objective function of the underwater robot, n is the total number of time steps, vi (t) is the actual speed of the i-th underwater robot at time t, is the predicted velocity of the i-th underwater robot at time t, q is the current posture quaternion, q d is the desired pose quaternion, λ is the weight coefficient, and C is the waterline image credibility.

[0044] A further improvement of the technical solution of the present invention is that: in step S5, the edge detection algorithm is used to identify the edges in the pre-processed waterline image, and double threshold segmentation is used to separate the waterline in the image, which specifically includes the following steps:

[0045] S5.1. Calculate the gradient magnitude and direction of each pixel in the waterline image;

[0046] S5.2. Apply non-maximum suppression to refine edges;

[0047] S5.3. Apply dual threshold detection to determine real and potential edges;

[0048] S5.4. Determine edges by edge tracking and hysteresis threshold.

[0049] A system for water flow estimation and ship waterline monitoring, characterized by using the joint design method described in any one of claims 1 to 8, comprising a perception module, a propeller module, a control module and a wireless communication module located on an underwater robot.

[0050] A further improvement of the above technical solution of the present invention is that: the perception module includes an airborne camera for collecting the waterline of the ship on the water surface and a Doppler velocimeter for obtaining the water flow velocity; the thruster module includes an underwater thruster and a drive module for providing power; the control module includes a processing unit for data processing and model deployment and an inertial navigation positioning device for obtaining the position and posture of the underwater robot; the wireless communication module includes a wireless radio frequency communicator for communication between underwater robots and an electromagnetic communicator for communication between the shore-based control center and the underwater robot.

[0051] Due to the adoption of the above technical solution, the technical advancements achieved by the present invention include:

[0052] Traditional ship waterline monitoring methods can usually only monitor the waterline at a single fixed location, and are often limited by the water environment and the ship's berthing position, making it impossible to monitor the waterline in all directions. The maneuverability of underwater robots allows for monitoring the ship's waterline from all angles and can also be quickly deployed.

[0053] The existing technology does not jointly design the underwater robot's ship waterline monitoring and water flow estimation, resulting in competition for communication and computing resources. This solution proposes a joint design method for water flow estimation and ship waterline monitoring. According to steps S2 and S3 in the claims, the underwater robot collects and fuses data, and then in step S4 of the claims, the posture of the underwater robot is adjusted in real time through water flow estimation to improve resource utilization. According to the claims specification, step S5 can ensure reliable and accurate identification of the ship's waterline. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a workflow diagram of a joint design method for water flow estimation and ship waterline monitoring according to the present invention;

[0055] Figure 2 This is a schematic diagram of a single underwater robot working in a combined design method and system for water flow estimation and ship waterline monitoring according to the present invention;

[0056] Figure 3 It is a working diagram of multiple underwater robots in a joint design method and system for water flow estimation and ship waterline monitoring of the present invention.

[0057] In the attached figure: 1. Shore-based control center; 2. Underwater robot communication and power supply composite cable car; 3. Monitoring area; 4. Underwater robot; 5. Ship. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. In the following description, the description of the public structure and technology is omitted to avoid unnecessary confusion of the concept of the present invention.

[0059] The present invention provides a system for estimating water flow and monitoring the waterline of a vessel 5 . The system mainly includes a sensing module, a propeller module, a control module, and a wireless communication module arranged on an underwater robot 4 .

[0060] Specifically, the above-mentioned perception module includes an onboard camera for collecting the waterline of the ship 5 on the water surface and a Doppler velocimeter for obtaining the water flow velocity.

[0061] The propeller module includes an underwater propeller and a drive module for providing power.

[0062] The control module includes a microcomputer for processing data and an inertial navigation and positioning device for obtaining the four postures of the underwater robot.

[0063] The wireless communication module includes a wireless radio frequency communicator for communication between the underwater robots 4 and an electromagnetic communicator for communication between the shore-based control center 1 and the underwater robot 4 .

[0064] At the same time, the present invention also provides a joint design method for water flow estimation and ship waterline monitoring, which refers to Figure 1 、 Figure 2 and Figure 3 It can be seen that the following steps are included:

[0065] S1. The shore-based control center 1 initializes the underwater robot 4, selects the ship 5 to be monitored, and matches the corresponding number of underwater robots 4 according to the number of water gauges to be monitored. At the same time, the shore-based control center divides each underwater robot 4 into its own monitoring area 3 according to the monitoring task requirements; specifically, the monitoring target is set to the waterline of the ship 5, the underwater robot 4 collects images of the waterline of the ship 5, establishes a water gauge data set, trains the water gauge data set through a neural network to obtain a deep learning model and deploys the underwater robot 4 to the mission waters, and the underwater robot 4 performs the monitoring task according to the path and monitoring area 3 planned by the shore-based control center 1.

[0066] The water gauge dataset established above is used for training the deep learning model. The water gauges of the monitored ship 5 are drawn on both sides of the large ship 5, distributed at the bow, midship and stern, and are specifically expressed as follows:

[0067]

[0068] Among them, C is the ship's water gauge reading, H d is the height of the water gauge character spacing, F is the character reading closest to the waterline, H w It is the height from the waterline to the nearest character.

[0069] S2. After the underwater robot 4 enters the monitoring area 3, the water flow is measured by a Doppler velocimeter, and the water flow velocity and flow direction data are collected simultaneously within a predetermined time interval. The actual motion trajectory of the underwater robot 4 is compared with the path planned by the shore-based control center, and a deviation is established. The position, speed and posture of the underwater robot 4 are obtained through the inertial navigation positioning device.

[0070] Specifically, after entering the monitoring area, the underwater robot 4 measures the water flow using a Doppler velocimeter, collecting water flow velocity and direction data simultaneously within a predetermined time interval. During the preset time interval, the Doppler velocimeter emits sound wave pulses and receives sound waves reflected from the seabed or particles in the water. The speed and direction of the water flow are determined by calculating the changes in the sound wave frequency. The formula is expressed as follows:

[0071]

[0072] Where v is the water velocity, Δf is the Doppler frequency shift, c is the speed of sound waves in water, f0 is the frequency of the sound waves emitted by the Doppler velocimeter, and θ is the angle between the direction of sound wave emission and the direction of water flow.

[0073] After collecting the water flow velocity and direction data, the underwater robot 4 will compare this data with its actual motion trajectory and establish the trajectory deviation Δd:

[0074]

[0075] Where Δd is the trajectory deviation, (x a ,x p ) is the actual position coordinate of the underwater robot, (y a ,y p ) are the predetermined position coordinates of the underwater robot.

[0076] S3. Each underwater robot 4 communicates through broadcasting, sends and receives other local water flow velocity data, and integrates the current trajectory deviation and flow velocity data through Kalman filtering to predict the future flow velocity field; and uses the underwater robot 4 to collect waterline images in real time, and preprocesses the collected waterline images, and then calculates the waterline image credibility based on the preprocessed images.

[0077] Specifically, when performing the monitoring task, each underwater robot 4 will exchange its own local measured water flow velocity data in a broadcast manner through the wireless communication module, and use Kalman filtering to integrate the current flow velocity data and trajectory deviation. Assume that the flow velocity data collected by each underwater robot 4 is v i (t), where i represents the i-th underwater robot 4 and t represents time. Set the initial state estimate And covariance matrix P0, according to the dynamic model of underwater robot 4, predict the state and covariance of the next moment:

[0078]

[0079] Where k represents the time step, is the predicted state matrix at time step k, F k is the state transition matrix, is the optimal estimated state matrix at time step k-1, B k is the control matrix, u k is the control input matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k, P k-1 is the covariance at time step k-1, is the transpose of the state transfer matrix, Q k is the process noise covariance matrix.

[0080] When there is new measurement data z k When , update the state estimate:

[0081]

[0082] P k =(IK k H k )P k|k-1

[0083] Among them, K k is the Kalman gain matrix, P k|k-1 is the prediction covariance matrix at time step k, H k is the observation matrix, is the transpose of the observation matrix, R k is the observation noise covariance matrix, is the updated state estimate matrix at time step k, is the predicted state matrix at time step k, I is the identity matrix, and P k The updated covariance matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k.

[0084] For the collected waterline images, we first perform preprocessing to remove noise and enhance contrast to improve image quality. The specific process is as follows:

[0085] S3.1. Denoising the image: Images often contain noise. To reduce random noise and improve image clarity, a Gaussian kernel function is used to convolve the image and remove the Gaussian noise.

[0086] S3.2. Enhance contrast, improve the visual effect of the image, and make the image details clearer. Use histogram equalization as follows:

[0087]

[0088] Where r(k) is the grayscale of the input image, T(r(k)) is the cumulative probability of grayscale r(k), s(k) is the grayscale of the output image, k is the grayscale of the pixel in the image, and n j is the pixel value at gray level j, and n is the total pixel value.

[0089] S3.3. After image preprocessing, the waterline image credibility C is calculated:

[0090]

[0091] Among them, N e is the number of edge pixels detected, N t is the total number of pixels in the image, is the variance of the image noise.

[0092] S4. The underwater robot 4 collects its own position, speed and attitude data through the inertial navigation positioning device to estimate its current attitude, and corrects its own attitude through the waterline image credibility and water flow velocity field prediction as feedback signals; when the waterline image credibility reaches the threshold, the underwater robot 4 adjusts its attitude through the water flow velocity field prediction to minimize the attitude optimization function and achieve stable monitoring; if the waterline image credibility does not reach the threshold, the underwater robot 4 continues to search and adjust its attitude along the path planned by the shore-based control center until the waterline image credibility reaches the threshold.

[0093] Specifically, the underwater robot 4 collects position, speed, and posture data, as well as the waterline image credibility, and performs joint posture correction to make the waterline image credibility reach a threshold and achieve posture stability. The optimization function is expressed as follows:

[0094]

[0095] in, is the minimization objective function of the underwater robot 4, n is the total number of time steps, v i (t) is the actual speed of the i-th underwater robot 4 at time t, is the predicted velocity of the i-th underwater robot 4 at time t, q is the current posture quaternion, q d is the desired pose quaternion, λ is the weight coefficient, and C is the waterline image credibility.

[0096] S5. Use an edge detection algorithm to identify the edges in the preprocessed waterline image, and use double threshold segmentation to separate the waterline in the image; specifically, the underwater robot 4 uses a deep learning algorithm to identify the water gauge number and combines it with the separated waterline to obtain the position of the waterline on the water gauge, and then uses the water gauge reading algorithm to obtain the draft of the ship and upload it to the shore-based control center 1.

[0097] The specific steps include:

[0098] S5.1. Calculate the gradient magnitude and direction of each pixel in the waterline image;

[0099] S5.2. Apply non-maximum suppression to refine edges;

[0100] S5.3. Apply dual threshold detection to determine real and potential edges;

[0101] S5.4. Determine edges by edge tracking and hysteresis threshold.

[0102] The YOLOv5 algorithm is used for draft gauge character recognition. The recognized draft gauge characters and the separated ship waterline are combined with the draft gauge reading algorithm to obtain the ship's draft value.

[0103] S6. According to the monitoring task requirements, the underwater robot (4) determines whether the monitoring task is completed. If not, it continues to perform monitoring. If the task is completed, the shore-based control center (1) issues a command to control the underwater robot (4) to return.

[0104] In the above embodiments, the present invention provides a joint design method and system for water flow estimation and ship waterline monitoring. The maneuverability of the underwater robot in the present invention can monitor the ship's waterline at various angles and can also be rapidly deployed. At the same time, the present invention adjusts the posture of the underwater robot in real time through water flow estimation, thereby improving resource utilization and ensuring reliable and accurate identification of the ship's waterline.

[0105] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the concept and scope of the present invention. Any modifications and improvements made to the technical solution of the present invention by a person of ordinary skill in the art without departing from the design concept of the present invention shall fall within the scope of protection of the present invention. The technical content for which protection is sought in the present invention is fully set forth in the claims.

Claims

1. A joint design method for water flow estimation and ship waterline monitoring, characterized by: The following steps are included: S1, initialize the underwater robot (4), select the ship (5) to be monitored, match the corresponding number of underwater robots (4) according to the number of water gauges to be monitored, and the shore-based control center divides each underwater robot (4) into respective monitoring areas (3) according to the monitoring task requirements; specifically, set the monitoring target as the waterline of the ship (5), the underwater robot (4) collects images of the waterline of the ship (5), establishes a water gauge data set, trains the water gauge data set through a neural network to obtain a deep learning model and deploys the underwater robot (4) to the mission water area, and the underwater robot (4) performs the monitoring task according to the path and monitoring area (3) planned by the shore-based control center (1); S2, after the underwater robot (4) enters the monitoring area (3), the water flow is measured by a Doppler velocimeter, and the water flow velocity and flow direction data are collected simultaneously within a predetermined time interval. The actual motion trajectory of the underwater robot (4) is compared with the path planned by the shore-based control center, and a deviation is established. The position, speed and attitude of the underwater robot (4) are obtained by an inertial navigation positioning device; S3, each underwater robot (4) communicates through broadcasting, sends and receives other local water flow velocity data, and integrates the current trajectory deviation and flow velocity data through Kalman filtering to predict the future flow velocity field; and uses the underwater robot (4) to collect waterline images in real time, and pre-processes the collected waterline images, and then calculates the waterline image credibility based on the pre-processed images; S4, the underwater robot (4) acquires its own position, speed and attitude data through the inertial navigation positioning device to obtain its current attitude, and corrects its attitude through the waterline image credibility and the water flow velocity field prediction as feedback signals; when the waterline image credibility reaches a threshold, the underwater robot (4) adjusts its attitude through the water flow velocity field prediction to minimize the attitude optimization function and achieve stable monitoring; if the waterline image credibility does not reach the threshold, the underwater robot (4) continues to search and adjust its attitude along the path planned by the shore-based control center until the waterline image credibility reaches the threshold; S5. Using an edge detection algorithm to identify edges in the pre-processed waterline image, and using double threshold segmentation to separate the waterline in the image; specifically, the underwater robot (4) uses a deep learning algorithm to identify the water gauge number and combines it with the separated waterline to obtain the position of the waterline on the water gauge to obtain the draft depth of the ship (5), and uploads it to the shore-based control center (1); S6. According to the monitoring task requirements, the underwater robot (4) determines whether the monitoring task is completed. If not, it continues to perform monitoring. If the task is completed, the shore-based control center (1) issues a command to control the underwater robot (4) to return.

2. A joint design method for water flow estimation and ship waterline monitoring according to claim 1, characterized in that: In step S1, a water gauge dataset is established for training the deep learning model. The water gauge of the monitored ship (5) is a water gauge drawn on both sides of the large ship (5), distributed at the bow, midship and stern, and is specifically expressed as follows: Among them, C is the ship's water gauge reading, H d is the height of the water gauge character spacing, F is the character reading closest to the waterline, H w It is the height from the waterline to the nearest character.

3. The combined design method for water flow estimation and ship waterline monitoring according to claim 1 is characterized by: In step S2, the water flow is measured by a Doppler velocimeter, and the water flow velocity and flow direction data are collected simultaneously within a predetermined time interval, specifically, At preset time intervals, the Doppler velocimeter emits sound wave pulses and receives sound waves reflected from the seabed or particles in the water. The speed and direction of the water flow are determined by calculating the changes in the sound wave frequency. The formula is: Where v is the water velocity, Δf is the Doppler frequency shift, c is the speed of sound waves in water, f0 is the frequency of the sound waves emitted by the Doppler velocimeter, and θ is the angle between the direction of sound wave emission and the direction of water flow.

4. The combined design method for water flow estimation and ship waterline monitoring according to claim 1 is characterized by: In step S2, the actual running trajectory of the underwater robot (4) is compared with the path planned by the shore-based control center to establish a deviation, specifically, After collecting the water flow velocity and direction data, the underwater robot (4) will compare these data with its actual motion trajectory and establish the trajectory deviation Δd: Where Δd is the trajectory deviation, (x a ,x p ) is the actual position coordinate of the underwater robot (4), (y a ,y p ) are the predetermined position coordinates of the underwater robot (4).

5. The combined design method for water flow estimation and ship waterline monitoring according to claim 1 is characterized by: In step S3, each underwater robot (4) communicates by broadcasting, sends and receives other local water flow velocity data, and integrates the current trajectory deviation and flow velocity data through Kalman filtering to predict the future flow velocity field, specifically, Assume that the flow velocity data collected by each underwater robot (4) is v i (t), where i represents the i-th underwater robot (4) and t represents time; Setting the initial state estimate And covariance matrix P0, predict the state and covariance of the next moment according to the underwater robot dynamics model: Where k represents the time step, is the predicted state matrix at time step k, F k is the state transition matrix, is the optimal estimated state matrix at time step k-1, B k is the control matrix, u k is the control input matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k, P k-1 is the covariance at time step k-1, is the transpose of the state transfer matrix, Q k is the process noise covariance matrix; When there is new measurement data z k When , update the state estimate: P k =(I-K k H k )P k|k-1 Among them, K k is the Kalman gain matrix, P k|k-1 is the prediction covariance matrix at time step k, H k is the observation matrix, is the transpose of the observation matrix, R k is the observation noise covariance matrix, is the updated state estimate matrix at time step k, is the predicted state matrix at time step k, I is the identity matrix, and P k The updated covariance matrix at time step k, P k|k-1 is the prediction covariance matrix at time step k.

6. The combined design method for water flow estimation and ship waterline monitoring according to claim 1, characterized in that: In step S3, the collected waterline image is preprocessed, and then the credibility of the waterline image is calculated based on the preprocessed image, which includes the following steps: S3.

1. Image denoising: Images often contain noise. To reduce random noise and improve image clarity, we use the Gaussian kernel function to convolve the image and remove the Gaussian noise. S3.

2. Enhance contrast, improve the visual effect of the image, and make the image details clearer. Use histogram equalization as follows: Where r(k) is the grayscale of the input image, T(r(k)) is the cumulative probability of grayscale r(k), s(k) is the grayscale of the output image, k is the grayscale of the pixel in the image, and n j is the pixel value of gray level j, n is the total pixel value; S3.

3. After image preprocessing, the waterline image credibility C is calculated: Among them, N e is the number of edge pixels detected, N t is the total number of pixels in the image, is the variance of the image noise.

7. The combined design method for water flow estimation and ship waterline monitoring according to claim 1, characterized in that: In step S4, the underwater robot (4) collects its own position, speed and posture data, as well as the waterline image credibility, and performs joint posture correction to make the waterline image credibility reach a threshold and achieve posture stability. The optimization function is expressed as follows: in, is the minimization objective function of the underwater robot (4), n is the total number of time steps, v i (t) is the actual speed of the i-th underwater robot at time t, is the predicted velocity of the ith underwater robot (4) at time t, q is the current posture quaternion, q d is the desired pose quaternion, λ is the weight coefficient, and C is the waterline image credibility.

8. The combined design method for water flow estimation and ship waterline monitoring according to claim 1, characterized in that: In step S5, the edge detection algorithm is used to identify the edges in the pre-processed waterline image, and double threshold segmentation is used to separate the waterline in the image. Specifically, the following steps are included: S5.

1. Calculate the gradient magnitude and direction of each pixel in the waterline image; S5.

2. Apply non-maximum suppression to refine edges; S5.

3. Apply dual threshold detection to determine real and potential edges; S5.

4. Determine edges by edge tracking and hysteresis threshold.

9. A system for water flow estimation and ship waterline monitoring, characterized by: The joint design method according to any one of claims 1 to 8 comprises a perception module, a propeller module, a control module and a wireless communication module located on an underwater robot (4).

10. The system for water flow estimation and ship waterline monitoring according to claim 9, characterized in that: The perception module includes an onboard camera for collecting the waterline of a ship on the water surface and a Doppler velocimeter for obtaining the water flow velocity; the propeller module includes an underwater propeller for providing power and a drive module thereof; the control module includes a processing unit for data processing and model deployment and an inertial navigation positioning device for obtaining the position and posture of the underwater robot (4); the wireless communication module includes a wireless radio frequency communicator for communication between the underwater robots (4) and an electromagnetic communicator for communication between the shore-based control center (1) and the underwater robot (4).

Images