Intelligent robot fish driving device with path planning and underwater terrain recognition functions

Abstract

The invention relates to an intelligent robot fish driving device with path planning and underwater terrain recognition. The intelligent robot fish driving device is composed of an electric grid, an underwater robot and a remote control module. The electric grid comprises a water surface end and an underwater end which are connected through a flexible steel wire rope, and the water surface end is provided with a generator, high-voltage pulse equipment and a main control module; the underwater end is composed of a plurality of underwater robots, each underwater robot is provided with an angle sensor, a depth sensor, a GPS module and an infrared camera, the main control module receives and processes information from the water surface cameras and the underwater robots, the underwater environment is analyzed through an mmsegment semantic segmentation algorithm, and a result is sent to the remote control module. And the remote control module is responsible for receiving picture and path information and realizing state monitoring and manual path setting functions. The device integrates the functions of path planning, underwater terrain recognition and the like, and aims to improve the efficiency and the intelligent level of fish driving operation.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent fish herding, and relates to an intelligent robot fish herding device with path planning and underwater terrain recognition. Background Art

[0002] At present, both the traditional way of driving fish and the existing electric fence device for driving fish have relatively low efficiency and low intelligence. As an important tool for fishing, improving the working efficiency and technological level of the electric fence device for driving fish has significant significance for reducing the production cost of fisheries and improving production efficiency.

[0003] The existing electric fence fish-driving device uses a motor-driven system to drag a floating boat and an electric fence to drive the fish. However, the device is large in size, poor in flexibility, and cannot only adjust the intensity and frequency of the electric pulses. It is not only power-consuming, but also very easy to cause damage to the device. At the same time, the staff cannot observe the operation of the fish-driving device and conduct manual intervention based on actual conditions, and the degree of automation is low. In addition, the actual fishing scenes are relatively diverse. For example, the surrounding and underwater environment of the reservoir is complex, and it may be located in the woods, mountains, etc. The electric grid is easily entangled by underwater foreign objects such as branches, which can also cause the device to be damaged and scrapped. Summary of the invention

[0004] In view of this, the object of the present invention is to provide an intelligent robot fish-herding device with path planning and underwater terrain recognition.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] An intelligent robot fish-driving device with path planning and underwater terrain recognition comprises an electric fence, wherein the electric fence comprises a surface end floating on the water surface and an underwater end connected with a plurality of underwater robots traveling underwater; the surface end and the underwater end are connected by a plurality of rows of soft steel wire ropes; the surface end is provided with a generator, a high-voltage pulse device, a main control module, a surface camera and left and right thrusters; the underwater robot at the underwater end comprises a robot body, an angle sensor, a depth sensor, a GPS module and an infrared camera; and further comprises a remote control module;

[0007] The generator is used to supply power to each device module; the high-voltage pulse equipment is used to apply high-voltage pulse current to the soft steel wire rope; the surface camera is used to collect surface images; the robot body is used to drive the underwater end of the electric grid to move; the angle sensor is used to detect the posture information of the robot body; the depth sensor is used to detect the depth of the robot body; the infrared camera is used to collect the surrounding environment information of the robot body; the main control module is connected to the surface camera, left and right thrusters and the underwater robot, and is used to receive the image information, angle information, depth information and position information sent back by the surface camera and the underwater robot, and use the mmsegmentation semantic segmentation algorithm to calculate the underwater environment picture and send it to the remote main control module; the remote control module is used to receive the picture and path information sent by the main control module to realize status monitoring and manual control to set the path function.

[0008] Furthermore, the specific steps of the main control module using the mmsegmentation semantic segmentation algorithm to calculate the underwater environment picture are as follows:

[0009] A1: Input stage, the underwater robot captures real-time images of the underwater environment through an infrared camera;

[0010] A2: In the preprocessing stage, the captured image is subjected to noise reduction and image enhancement processing; the image enhancement uses histogram equalization, and its calculation formula is:

[0011]

[0012] Among them, I out (x,y) is the intensity value of the output image at point (x,y); I in (x,y) is the intensity value of the input image at point (x,y); L is the number of possible values ​​of pixel intensity; N is the total number of pixels in the image. H(k) is the cumulative histogram, which represents the number of pixels with intensity value k or less;

[0013] A3: semantic segmentation stage, using mmsegmentation semantic segmentation algorithm to perform semantic segmentation on the preprocessed image and divide the underwater environment in the image into different areas;

[0014] A4: Analysis phase: Based on the semantic segmentation results, obstacles in the underwater environment are identified and safe areas are marked;

[0015] A5: Output stage: output the semantic segmentation results and obstacle location information to the main control module and transmit them to the remote control module;

[0016] A6: During the obstacle avoidance phase, according to the output obstacle position information, the underwater robot body is controlled to move the underwater end of the electric fence and the left and right thrusters are controlled to move the surface end of the electric fence to avoid obstacles.

[0017] Furthermore, the semantic segmentation stage in step A3 specifically includes the following steps:

[0018] A31: Build a deep learning model and select a network architecture suitable for underwater image segmentation;

[0019] A32: Training model: Use the labeled underwater image dataset to train the deep learning model until the model achieves the preset segmentation effect and the numerical performance of relevant evaluation indicators meets expectations;

[0020] A33: Model optimization: pruning and quantization of the trained model to adapt to the computing resources of the underwater robot.

[0021] Furthermore, in step A6, the method for controlling the underwater robot to avoid obstacles is: obtaining obstacle information according to semantic segmentation, providing movement parameters of the underwater robot in combination with a depth sensor and an angle sensor, and then providing a path to avoid obstacles.

[0022] Furthermore, the main control module also has a dynamic path planning function, including the following steps:

[0023] B1: Input stage, the main control module receives the current position information of the underwater robot provided by the GPS module;

[0024] B2: Data processing stage, the main control module analyzes historical path data and identifies the paths corresponding to areas with better underwater environmental conditions;

[0025] B3: Path planning stage, the main control module uses a dynamic programming algorithm to combine the current position information and the optimal solution of the local underwater environment to comprehensively calculate the optimal path;

[0026] B4: Output stage, the main control module outputs the calculated optimal path to the robot body;

[0027] B5: Execution phase: the robot body moves along the optimal path.

[0028] Furthermore, in step B3, the path planning stage specifically includes the following sub-steps:

[0029] B31: Define states, divide the underwater environment into several states, each state represents a possible location;

[0030] B32: Define transition, determine the rules and cost of transitioning from one state to another: define a cost function C(s,a,s′), which represents the cost of transitioning from state s to state s′ through action a;

[0031] B33: Calculate the optimal path. Starting from the initial state, find the path with the minimum cost through iterative calculation, that is, the optimal path. The mathematical formula for calculation is:

[0032]

[0033] Among them, V(s) is the value function of state s, which represents the minimum cost of reaching the target state from state s; A is the set of available actions, which represents all possible actions that can be taken from the current state s; C(s,a,s ′ ) is a transition from state s to state s through action a ′ is the cost function; γ is the discount factor used to measure the current value of future rewards; V(s ′ ) is the state s ′ The value function represents the value of ′ The minimum cost to start reaching the goal state.

[0034] Furthermore, the middle and lower parts of the soft steel wire rope are fixed by glue sticks, thereby dividing the soft steel wire rope into three sections.

[0035] Furthermore, the infrared camera is provided with an LED light for fill light.

[0036] The beneficial effects of the present invention are as follows: the present invention designs a highly intelligent, good fish-driving effect, good safety performance, low operating cost, and convenient monitoring and intervention intelligent electric fence and underwater robot collaborative fish-driving device, which uses deep learning, semantic segmentation, GPS path planning navigation, PID control and other algorithms to solve the problems of weak ability of the device to cope with complex underwater environments, low degree of automation, device monitoring and maintenance, and device safety, which can promote the high-quality development of fisheries and has high practicality and promotion value. Specifically:

[0037] (1) The present invention performs semantic segmentation through the mmsegmentation algorithm, which can accurately identify obstacles in the underwater environment and improve the accuracy of obstacle avoidance.

[0038] (2) The present invention utilizes a dynamic programming algorithm combined with GPS module positioning to dynamically calculate the optimal path and improve the efficiency of driving fish.

[0039] (3) The present invention can adapt to complex underwater environments and has strong practicality and promotion value.

[0040] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0042] Figure 1 This is a structural diagram of an intelligent robot fish-herding device with path planning and underwater terrain recognition;

[0043] Figure 2 It is a schematic diagram of system connection relationship;

[0044] Figure 3 It is a schematic diagram of the connection relationship of device modules;

[0045] Figure numerals: generator 1, high-voltage pulse equipment 2, main control module 3, surface camera 4, buoy 5, left and right thrusters 6, soft steel wire rope 7, glue stick 8, underwater robot body 9, depth sensor 10, infrared camera 11, remote control module 12. DETAILED DESCRIPTION

[0046] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0047] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0048] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0049] The present invention provides an intelligent robot fish-driving device with path planning and underwater terrain recognition, which consists of a remote control module, a main control module and an underwater robot module. The remote control module consists of a touch display screen and a WIFI module, which is wirelessly connected to the main control module through multi-level communication, receives information sent back by the main control module in real time, and allows staff to monitor the working status of the device in time and perform manual intervention in time; the main control module integrates functional modules such as a power module, a WIFI module, and a camera, and is connected to the underwater robot module through serial port communication, outputs a surface video stream according to a surface camera acquisition picture, calculates and outputs a semantic segmentation video stream of the underwater environment according to an infrared camera acquisition picture, and performs SLAM underwater environment mapping when entering a new water area; the underwater robot module consists of a robot controller composed of an infrared camera, a depth sensor, and an angle sensor, and an underwater robot body, which can detect the posture of the underwater robot, the depth of the device, and receive the obstacle avoidance command sent back by the main control module according to the semantic segmentation video stream to control the device, so as to achieve underwater obstacle avoidance.

[0050] The working principle of the device is: the underwater robot starts its propeller to break the current state of balance between its gravity and the buoyancy of the float and dives to a certain depth, then turns on the electric fence, and the infrared camera detects the target, so that the device in the working state moves towards the school of fish. To prevent the device from touching underwater foreign objects, the depth sensor integrated in the underwater robot detects the current water depth in real time. If the underwater depth is reached, the underwater robot will push the device to float up in time, and use the PID algorithm to control the water depth of the underwater robot within a certain range. In response to the problem of complex underwater terrain, the underwater robot uses the infrared camera to send the picture back to the main control module for semantic segmentation of the underwater environment. When a relatively high land mass appears on the bottom of the water, the device is slowed down and floated at the same time, finally realizing flexible obstacle avoidance under the water; to avoid short circuit caused by the end electric fence being too close, the underwater robot regularly measures the distance between the electric fences. When the electric fence distance reaches the touch critical value, the underwater robot drives the float forward to tighten the electric fence until it stops tightening after a safe distance. When working for the first time, the device will automatically operate in the water area according to the water area boundary fed back by GPS, providing a reference for path planning; the device can also be combined with the GPS positioning module. After the remote control module sets the travel path, the device will run according to the established path; at the same time, the device's travel path will also be fed back to the remote control module in real time, which is convenient for the route planning of the next operation. After multiple operations, the system will record the travel paths of the device and use the dynamic programming algorithm to give the optimal solution for the fish-driving path; in order to solve the problem that the existing fish-driving device cannot monitor the operation of the device in real time, the device can also feed back the video stream of the surface camera and the underwater environment video stream output by the infrared camera after the semantic segmentation algorithm is run to the remote control module, which is convenient for real-time control of the device and timely human intervention.

[0051] like Figure 1 As shown, the device specifically includes the following structure:

[0052] Generator 1: used to supply power to the entire device including electric grid, electric propulsion, underwater robot and other modules. Considering that the high-voltage pulse equipment requires about 300W of power and other power-consuming equipment requires about 300W of power, the entire system requires about 600W of power. It is more appropriate to use a generator with a power consumption of less than 1KW, such as Honda EU1000i.

[0053] High-voltage pulse equipment 2: input 220V voltage to generate 300V-500V high-voltage pulses, and the positive and negative poles of the high-voltage pulses are added to the four soft steel wire ropes 7. The equipment can use Tester voltage pulse generator and the like.

[0054] Main control module 3: It adopts embedded system, integrates power module, WIFI module, GPS module, etc., is connected to the surface camera 4, and transmits the image of the surface camera 4 and the device location in real time; at the same time, it runs the deep learning mmsegmentation semantic segmentation algorithm, connects to the underwater robot through the serial port, inputs the underwater robot infrared camera image and other information for comprehensive analysis, outputs the underwater environment semantic segmentation result image, and realizes the underwater obstacle avoidance function. The embedded system can use JETSON NANO, the WIFI module can use the relevant module of LiChuang, and the GPS module can use Shenzhen Hefangyuan's GU620.

[0055] Surface camera 4: connected to the main control module, used to monitor the water surface conditions.

[0056] Float 5: makes the main control module of the device and the underwater robot in the initial state float on the water surface. The float on the surface end is used to place and protect the main control module. The float on the underwater end allows the underwater robot module to use the water surface as a reference, break the current balance of gravity and buoyancy and start to float down, and continuously correct the working posture of the electric grid.

[0057] Left and right thrusters 6: The propulsion device moves on the water surface and can also turn by using differential speed. BlueRobotics' T200 Thruster thruster can be used.

[0058] Electric fence and connecting cable: The electric fence is composed of soft steel wire rope 7, which is divided into positive and negative poles. The positive and negative poles of the electric fence are alternately distributed from left to right. An electric field will be formed between the positive and negative poles of the electric fence, which will stimulate the fish. The length of the electric fence can be determined according to the actual situation of the water area. For example, in the context of a reservoir, the water depth of 15 meters can be set to 6 meters in the upper section, 7 meters in the middle section, and 2 meters in the lower section according to two rubber sticks; the water depth of 10 meters can be designed as 4 meters in the upper section, 5 meters in the middle section, and 1 meter in the lower section. The connecting cable is responsible for communicating between the main control module and the underwater robot module, which is similar to the routing of the electric fence. The electric fence can use products such as Kencove, and the connecting cable can use products such as Oriental Cable Group.

[0059] Glue stick 8: Two glue sticks are set in the middle of the electric grid to prevent the electric grid from short-circuiting on the one hand, and to partially fix the routing of the electric grid and the connecting cable on the other hand to prevent the electric grid or the connecting cable from being entangled during the movement of the underwater robot.

[0060] Underwater robot: including robot body 9, angle sensor, depth sensor 10, GPS module and infrared camera 11, etc., connected to the main control module to transmit parameters in real time to realize the function of electric fence obstacle avoidance control. The angle sensor can adopt PCF8591 rotation angle sensor. Depth sensor 10 is used to measure the depth of the underwater robot (usually also the depth of the end of the electric fence device) to provide depth parameter support for electric fence obstacle avoidance. The depth sensor can adopt Shenzhen Huace HC-200 depth sensor. Infrared camera 11: used to output underwater environment video stream. The underwater environment is usually dim and turbid, which is not easy to identify clearly. Therefore, the solution of infrared camera plus LED fill light is adopted. The infrared camera can adopt Hikvision DS-2CD6886G0-CO (8mm).

[0061] Remote control module 12: It is composed of a touch screen and a WIFI module powered by a lithium battery module. It receives the images, paths and other information sent back by the main control module via WIFI in real time, so as to facilitate the functions of device operation status monitoring, path planning, manual intervention and the like.

[0062] like Figure 2 The figure shows the system connection relationship diagram. The device is mainly composed of three core modules: remote control module, main control module and underwater robot module. The main control module is also responsible for controlling the operation of the pulse fish driving module. Figure 3 As shown in the figure, the remote control module is mainly composed of a lithium battery module, a touch screen and a WIFI module. The lithium battery module is responsible for power supply, the WIFI module is responsible for communication with the main control module, and the touch screen is responsible for visual interaction. The main control module consists of six parts: power module, main control controller (i.e., embedded system), WIFI module, GPS module, camera and left and right underwater propulsion. It is also connected to the pulse fish-driving module. The pulse fish-driving module includes high-voltage pulse equipment and electric grids to complete the working voltage conversion and operation of the electric grid; the main control controller is connected to the robot controller, responsible for using the model algorithm for calculation and reasoning, and outputting appropriate signals to related equipment; the WIFI module is responsible for establishing communication with the remote control module; the GPS module is responsible for real-time positioning of the device, helping the visualization device to run the path, and providing a reference for staff decision-making and dynamic planning of the optimal path.

[0063] The underwater robot module consists of an underwater robot body and a robot controller that integrates an LED infrared camera, an angle sensor, and a depth sensor. The LED infrared camera is responsible for capturing images when the underwater light is dim and the water quality is complex; the angle sensor is responsible for detecting the current posture of the underwater robot so that the robot always maintains a forward posture; the depth sensor is responsible for detecting the current depth of the underwater robot (the underwater robot is usually close to the depth of the electric fence end), and combined with the infrared camera's semantic segmentation of the underwater situation, the robot body promptly activates the thruster to drive the device to float up and keep the device in a safe water depth.

[0064] The power module of the main control module is turned on, and then the remote control module is turned on. The surface camera and the infrared camera of the underwater robot module return the video stream. If the video stream shows that the device is in normal state, the propeller of the underwater robot body starts to work, breaking the buoyancy of the current device end buoy and the device weight balance state to start diving. After reaching the lower limit of the predetermined safe water depth, the propeller of the underwater robot body and the left and right propellers controlled by the main control module start to operate at the same speed, pushing the entire device forward along the path set by the remote control module. If a turn is required, the left and right propellers can be made to produce different speeds for differential turning. At the same time, the underwater robot also adjusts the propeller speed in time to keep it consistent with the left and right underwater propellers. At the same time, the environment semantic segmentation, depth monitoring and PID control algorithms in the main control module start to run, so that the left and right propellers and the underwater robot body propellers produce the same speed as much as possible, so that the device can operate in a suitable posture; at the same time, the underwater environment, the depth of the device, the working condition of the electric fence, etc. are monitored in real time to ensure the normal operation of the device. The staff uses the remote control module to observe the operation of the device in real time through the above-water and underwater cameras, and combines the GPS module to monitor whether the device's running track is normal and whether it is working within the water area, so as to switch the automatic / manual mode in time to intervene in emergencies. When the device completes the operation according to the established path, it stops working automatically.

[0065] In addition, when the underwater robot detects that the distance between the end electric grids has reached a dangerous distance, the underwater robot is controlled to accelerate, thereby driving the buoy at the end of the electric grid to move forward and tighten the electric grid, thereby increasing the distance between the electric grids until the distance between the electric grids remains within a safe distance for a certain period of time. The PID control algorithm is then used to slowly restore the underwater robot to the same speed as the left and right underwater thrusters.

[0066] The underwater robot module's infrared camera with LED imaging is clear and bright under LED fill light, and can effectively deal with situations such as turbid river water and high sediment content. When the river water is too turbid and the collected video signal is blurred, the device's safe water depth range is increased by a certain value, allowing the device to operate according to the given optimal path.

[0067] The steps of the above-mentioned mmsegmentation semantic segmentation algorithm for underwater robot obstacle avoidance are as follows:

[0068] A1: Input stage, the underwater robot captures real-time images of the underwater environment through an infrared camera;

[0069] A2: In the preprocessing stage, the captured image is subjected to noise reduction and image enhancement processing; the image enhancement uses histogram equalization, and its calculation formula is:

[0070]

[0071] Among them, I out (x,y) is the intensity value of the output image at point (x,y); I in (x,y) is the intensity value of the input image at point (x,y); L is the number of possible values ​​of pixel intensity; N is the total number of pixels in the image. H(k) is the cumulative histogram, which represents the number of pixels with intensity value k or less;

[0072] A3: Semantic segmentation stage: The mmsegmentation semantic segmentation algorithm is used to perform semantic segmentation on the preprocessed image and divide the underwater environment in the image into different areas. Specifically, it includes:

[0073] A31: Build a deep learning model and select a network architecture suitable for underwater image segmentation;

[0074] A32: Training model: Use the labeled underwater image dataset to train the deep learning model until the model achieves the preset segmentation effect and the numerical performance of relevant evaluation indicators meets expectations;

[0075] A33: Model optimization, pruning and quantizing the trained model to adapt to the computing resources of the underwater robot;

[0076] A4: Analysis phase: Based on the semantic segmentation results, obstacles in the underwater environment are identified and safe areas are marked;

[0077] A5: Output stage: output the semantic segmentation results and obstacle location information to the main control module and transmit them to the remote control module;

[0078] A6: In the obstacle avoidance stage, according to the output obstacle position information, the underwater robot body is controlled to move the underwater end of the electric grid and the left and right thrusters are controlled to move the surface end of the electric grid to avoid obstacles. Specifically, the obstacle information is obtained based on semantic segmentation, and the underwater robot movement parameters are given in combination with the depth sensor and angle sensor, thereby giving a path to avoid obstacles.

[0079] The main control module also has a dynamic path planning function, including the following steps:

[0080] B1: Input stage, the main control module receives the current position information of the underwater robot provided by the GPS module;

[0081] B2: Data processing stage, the main control module analyzes historical path data and identifies the paths corresponding to areas with better underwater environmental conditions;

[0082] B3: Path planning stage: The main control module uses a dynamic programming algorithm to combine the current location information and the optimal solution of the local underwater environment to comprehensively calculate the optimal path; specifically, it includes the following sub-steps:

[0083] B31: Define states, divide the underwater environment into several states, each state represents a possible location;

[0084] B32: Define transition, determine the rules and cost of transitioning from one state to another: define a cost function C(s,a,s′), which represents the cost of transitioning from state s to state s′ through action a;

[0085] B33: Calculate the optimal path. Starting from the initial state, find the path with the minimum cost through iterative calculation, that is, the optimal path. The mathematical formula for calculation is:

[0086]

[0087] Among them, V(s) is the value function of state s, which represents the minimum cost of reaching the target state from state s; A is the set of available actions, which represents all possible actions that can be taken from the current state s; C(s,a,s ′ ) is a transition from state s to state s through action a ′ is the cost function; γ is the discount factor used to measure the current value of future rewards; V(s ′ ) is the state s ′ The value function represents the value of ′ The minimum cost to start reaching the goal state;

[0088] B4: Output stage, the main control module outputs the calculated optimal path to the robot body;

[0089] B5: Execution phase: the robot body moves along the optimal path.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. An intelligent robot fish-herding device with path planning and underwater terrain recognition, characterized in that: The electric fence includes a surface end floating on the water surface and an underwater end connected with a plurality of underwater robots traveling underwater; the surface end and the underwater end are connected by a plurality of rows of soft steel wire ropes; the surface end is provided with a generator, a high-voltage pulse device, a main control module, a surface camera and left and right thrusters; the underwater robot at the underwater end includes a robot body, an angle sensor, a depth sensor, a GPS module and an infrared camera; and also includes a remote control module; The generator is used to supply power to each device module; the high-voltage pulse equipment is used to apply high-voltage pulse current to the soft steel wire rope; the surface camera is used to collect surface images; the robot body is used to drive the underwater end of the electric grid to move; the angle sensor is used to detect the posture information of the robot body; the depth sensor is used to detect the depth of the robot body; the infrared camera is used to collect the surrounding environment information of the robot body; the main control module is connected to the surface camera, left and right thrusters and the underwater robot, and is used to receive the image information, angle information, depth information and position information sent back by the surface camera and the underwater robot, and use the mmsegmentation semantic segmentation algorithm to calculate the underwater environment picture and send it to the remote main control module; the remote control module is used to receive the picture and path information sent by the main control module to realize status monitoring and manual control to set the path function.

2. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 1 is characterized in that: The specific steps of the main control module using the mmsegmentation semantic segmentation algorithm to calculate the underwater environment picture are as follows: A1: Input stage, the underwater robot captures real-time images of the underwater environment through an infrared camera; A2: In the preprocessing stage, the captured image is subjected to noise reduction and image enhancement processing; the image enhancement uses histogram equalization, and its calculation formula is: Among them, I out (x,y) is the intensity value of the output image at point (x,y); I in (x,y) is the intensity value of the input image at point (x,y); L is the number of possible values ​​of pixel intensity; N is the total number of pixels in the image. H(k) is the cumulative histogram, which represents the number of pixels with intensity value k or less; A3: semantic segmentation stage, using mmsegmentation semantic segmentation algorithm to perform semantic segmentation on the preprocessed image and divide the underwater environment in the image into different areas; A4: Analysis phase: Based on the semantic segmentation results, obstacles in the underwater environment are identified and safe areas are marked; A5: Output stage: output the semantic segmentation results and obstacle location information to the main control module and transmit them to the remote control module; A6: During the obstacle avoidance phase, according to the output obstacle position information, the underwater robot body is controlled to move the underwater end of the electric fence and the left and right thrusters are controlled to move the surface end of the electric fence to avoid obstacles.

3. The intelligent robot fish-driving device with path planning and underwater terrain recognition according to claim 2 is characterized in that: The semantic segmentation stage described in step A3 specifically includes the following steps: A31: Build a deep learning model and select a network architecture suitable for underwater image segmentation; A32: Training model: Use the labeled underwater image dataset to train the deep learning model until the model achieves the preset segmentation effect and the numerical performance of relevant evaluation indicators meets expectations; A33: Model optimization: pruning and quantization of the trained model to adapt to the computing resources of the underwater robot.

4. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 2 is characterized in that: In step A6, the method for controlling the underwater robot to avoid obstacles is: obtaining obstacle information according to semantic segmentation, providing movement parameters of the underwater robot in combination with a depth sensor and an angle sensor, and then providing a path to avoid obstacles.

5. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 1 is characterized in that: The main control module also has a dynamic path planning function, including the following steps: B1: Input stage, the main control module receives the current position information of the underwater robot provided by the GPS module; B2: Data processing stage, the main control module analyzes historical path data and identifies the paths corresponding to areas with better underwater environmental conditions; B3: Path planning stage, the main control module uses a dynamic programming algorithm to combine the current position information and the optimal solution of the local underwater environment to comprehensively calculate the optimal path; B4: Output stage, the main control module outputs the calculated optimal path to the robot body; B5: Execution phase: the robot body moves along the optimal path.

6. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 5 is characterized in that: In step B3, the path planning stage specifically includes the following sub-steps: B31: Define states, divide the underwater environment into several states, each state represents a possible location; B32: Define transition, determine the rules and cost of transitioning from one state to another: define a cost function C(s,a,s′), which represents the cost of transitioning from state s to state s′ through action a; B33: Calculate the optimal path. Starting from the initial state, find the path with the minimum cost through iterative calculation, that is, the optimal path. The mathematical formula for calculation is: Among them, V(s) is the value function of state s, which represents the minimum cost of reaching the target state from state s; A is the set of available actions, which represents all possible actions that can be taken from the current state s; C(s,a,s ′ ) is a transition from state s to state s through action a ′ is the cost function; γ is the discount factor used to measure the current value of future rewards; V(s ′ ) is the state s ′ The value function represents the value of ′ The minimum cost to start reaching the goal state.

7. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 1 is characterized in that: The middle and lower parts of the soft steel wire rope are fixed by glue sticks, thereby dividing the soft steel wire rope into three sections.

8. The intelligent robot fish-herding device with path planning and underwater terrain recognition according to claim 1 is characterized in that: The infrared camera is provided with an LED light for supplementary lighting.

Images