Intelligent killing equipment for long-spiny starfishes and control method of intelligent killing equipment

Abstract

The invention relates to the technical field of marine organism treatment, in particular to an intelligent killing device for long-spiny starfishes and a control method of the intelligent killing device. The equipment is composed of a platform module, an injection module and an intelligent identification module, the platform module is made of a high-strength corrosion-resistant material, is equipped with a multi-direction electric propeller, a high-definition holder camera and a high-performance control module, has the functions of automatic depth setting, navigation setting, attitude stabilization and the like, and can flexibly move and keep operation stable under complex sea conditions. The injection module is composed of an agent storage tank, an injector and a high-penetrating-power needle head, the agent injection amount and speed can be automatically adjusted according to the target position, size and posture, and accurate and efficient killing is achieved. The intelligent recognition module recognizes and positions the morphological characteristics of the long spiny starfish in real time based on a deep learning algorithm and a high-definition image sensor, and ensures the accuracy and automation of the operation process. According to the invention, long-time underwater autonomous killing operation is realized, the killing efficiency of the long-spiny starfishes is remarkably improved, and the risk of manual diving is reduced.

Description

Technical Field

[0001] This invention relates to the field of marine biological control technology, and in particular to an intelligent extermination device for crown-of-thorns starfish and its control method. Background Technology

[0002] Crown-of-thorns starfish is a significant threat to coral reef ecosystems. Its rapid proliferation quickly consumes coral tissue, causing widespread coral death and disrupting the marine ecological balance. Current control methods primarily rely on manual divers injecting toxic agents or mechanical harvesting. However, these methods have several drawbacks. First, manual operations are labor-intensive and inefficient, making sustained eradication efforts difficult over large areas. Second, diving operations pose high personal safety risks, especially in deep water or complex sea conditions. Third, manual injection struggles to ensure accurate dosage and injection site, often resulting in wasted agents or incomplete eradication. Furthermore, some traditional identification equipment has limited capabilities, failing to accurately distinguish crown-of-thorns starfish in turbid or multi-species environments, leading to misidentification or missed eradication. Existing technologies have significant shortcomings in intelligent identification, automated operation, precise injection, and long-term underwater operational stability. Therefore, there is a need to develop an intelligent eradication device with autonomous identification, precise positioning, and automatic injection capabilities to achieve more efficient, safe, and environmentally friendly intelligent eradication and control of crown-of-thorns starfish. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides an intelligent extermination device for crown-of-thorns starfish and its control method.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides an intelligent extermination device for crown-of-thorns starfish, the device comprising a platform module, an injection module, and an intelligent identification module: The platform module includes a platform frame made of compressed polypropylene material, and an outer shell made of 6061-T6 hard anodized material. The platform frame is designed to be streamlined to reduce resistance during movement in water. The maximum diving depth is 50m. The platform frame has multiple functional installation areas, and each functional installation area is used to install a sub-functional module. The injection module includes a drug storage tank, a syringe, an injection needle, a delivery tube, and connectors. The drug storage tank is made of corrosion-resistant and well-sealed polytetrafluoroethylene material, which can resist the corrosion of different chemical agents. The maximum drug carrying capacity of the drug storage tank is 500ml, and the injection volume at one time is 5ml to 10ml. The syringe includes a drug-propelling piston connected to a drive motor, which controls the injection action and stroke of the drug-propelling piston. The delivery tube and connectors are made of medical-grade silicone material, which has good flexibility and chemical stability. The intelligent recognition module integrates deep learning algorithms and image recognition technology to perform biological morphological feature recognition on captured images of crown-of-thorns starfish.

[0005] Furthermore, in a preferred embodiment of the present invention, the sub-functional module includes a power mechanism, a camera mechanism, an energy mechanism, a control module, an injection module, and an intelligent recognition module.

[0006] Furthermore, in a preferred embodiment of the present invention, the power mechanism is provided with eight electric thrusters, wherein four electric thrusters are horizontally installed and four electric thrusters are vertically installed and distributed at different positions on the platform frame to achieve multi-directional thrust output, generating a forward thrust of 24 kgf, a lateral thrust of 14 kgf, and a vertical thrust of 28 kgf. Each electric thruster integrates a propulsion motor, which is a high-torque, low-speed waterproof motor model and is externally equipped with a waterproof protective shell. The power mechanism is also equipped with a speed sensor and a steering angle sensor.

[0007] Furthermore, in a preferred embodiment of the present invention, the camera mechanism is provided with multiple high-definition PTZ cameras, which are distributed at the front end and bottom of the platform frame. Each high-definition PTZ camera is provided with sealing material and protective coating on its exterior. The camera mechanism is connected to the control module and the intelligent recognition module through a high-speed data transmission line.

[0008] Furthermore, in a preferred embodiment of the present invention, the energy mechanism integrates a rechargeable battery pack, which is equipped with a power management system.

[0009] Furthermore, in a preferred embodiment of the present invention, the control module integrates a microprocessor, which connects to and coordinates the control of the power mechanism, camera mechanism, energy mechanism and injection module. The control module is provided with a manual mode and an automatic mode. The manual mode is used for equipment debugging and intervention operations in special circumstances, while the automatic mode is used as the main control mode for the equipment during normal operation.

[0010] The second aspect of this invention provides a control method for an intelligent starfish eradication device, applicable to any of the intelligent starfish eradication devices described in the invention, specifically comprising the following steps: By combining a high-definition gimbal camera and image sensor with an intelligent recognition module, the target sea area is scanned in all directions to obtain underwater image recognition results of crown-of-thorns starfish in the target sea area. The current morphological point cloud of crown-of-thorns starfish in the target sea area is obtained by underwater image recognition results. Taking the crown-of-thorns starfish intelligent extermination device as the field center, the pose of the current morphological point cloud is semantically matched and registered based on the morphological semantic labels of crown-of-thorns starfish under different phenotypic posture conditions to determine the real-time position and current posture of crown-of-thorns starfish. Based on real-time location control, the intelligent starfish eradication device moves to the vicinity of the starfish under the drive of the power mechanism, and controls the injection module to aim at the injection site of the starfish based on the current posture. Obtain the instructions for use of the pesticide, and extract from the instructions for use several preset growth factors of the pesticide against crown of thorns star, the predetermined value of each preset growth factor, and the prescribed baseline dose for different combinations of preset growth factors. Extract the actual element values ​​of crown of thorns star on each preset growth element. Based on the fitting of the predetermined element values ​​and the actual element values, construct a radar model of the predetermined growth element and a radar model of the actual growth element. Based on the breakthrough-stagnation trend of the indicators after the actual growth element radar model is registered with the radar model of the predetermined growth element, perform a collaborative split on the drug application weight ratio of the preset growth element to obtain a drug application collaborative tree for each preset growth element. The leaf node gain contribution during the splitting process of each preset growth element is calculated by tracking the drug application synergy tree to obtain the actual drug application contribution. The actual drug application contribution of each preset growth element is accumulated and integrated to obtain the actual killing dose of the killing agent applied to the target sea area for starfish. The injection module controls the actual killing dose to inject the killing agent into the crown-of-thorns starfish at an appropriate speed and angle.

[0011] Furthermore, in a preferred embodiment of the present invention, the step of obtaining the current morphological point cloud of crown-of-thorns starfish in the target sea area through underwater image recognition results, taking the crown-of-thorns starfish intelligent extermination device as the field center, and performing semantic matching registration transformation on the pose of the current morphological point cloud based on the morphological semantic labels of crown-of-thorns starfish under different phenotypic posture conditions to determine the real-time position and current posture of crown-of-thorns starfish, specifically includes the following steps: Based on the underwater image recognition results, extract the current morphological point cloud of the crown-of-thorns starfish in the target sea area from the real-time image frame, and at the same time obtain the marine biological knowledge graph based on big data; A semantic segmentation network is constructed, and morphological semantic labels of crown-of-thorns starfish under different phenotypic postures are obtained by retrieving marine biological knowledge graphs. Based on the morphological semantic labels, the current morphological point cloud is formulated in the semantic segmentation network to obtain the morphological semantic terms of each current morphological point cloud. Establish a Manhattan matching matrix, use the Manhattan matching matrix to calculate the Manhattan matching distance of semantic categories between semantic terms of each morphological semantic term, obtain the category matching degree of each point cloud semantic, if the category matching degree is greater than the preset category matching degree, then aggregate the current morphological point cloud corresponding to the semantic annotation term of that morphological term to form a point cloud nearest neighbor matching cluster. Define the current morphological point cloud as the source morphological point cloud, introduce the softmax probability algorithm to calculate the softmax probability distance between the source morphological point cloud and each nearest neighbor point cloud in the nearest neighbor matching cluster, extract only the nearest neighbor point cloud corresponding to the minimum softmax probability distance, label it as the nearest neighbor point cloud, and determine the semantic entropy weight of the nearest neighbor point cloud according to the softmax probability distance, so as to obtain several sets of source morphological point cloud-nearest neighbor point cloud matching point pairs. The LM algorithm is introduced. Based on semantic entropy weight, the residual and Jacobian matrix of each source morphological point cloud-nearest neighbor point cloud matching point pair are calculated in the LM algorithm. The update step size and registration error are obtained respectively. The optimal registration transformation of the current semantic weighted iteration is generated by minimizing the iterative registration error according to the update step size. Based on the current semantic weighted iteration, the optimal registration transformation is used to align and register the source morphological point cloud with the location of the crown-of-thorns starfish intelligent extermination device as the field center. Finally, the pose registration distribution field is output, and the real-time position and current attitude of the crown-of-thorns starfish are determined according to the pose registration distribution field.

[0012] Furthermore, in a preferred embodiment of the present invention, the step of extracting the actual element values ​​of crown-of-thorns starfish on each preset growth element, constructing a radar model of the preset growth elements and a radar model of the actual growth elements based on the fitting of the predetermined element values ​​and the actual element values, and performing synergistic splitting of the drug application weight ratio of the preset growth elements based on the breakthrough-retention trend of the indicators after the actual growth element radar model is registered with the predetermined growth element radar model to obtain a drug application synergistic tree for each preset growth element, specifically includes the following steps: Create a radar framework, and extend multiple extermination index radar branches dominated by preset growth elements from the center of the radar framework. Fit the predetermined element values ​​to the corresponding extermination index radar branches in the radar framework to generate a radar model of predetermined growth elements for the extermination agent for crown-of-thorns starfish. The actual element values ​​of crown-of-thorns starfish in the target sea area are extracted by underwater image recognition results. The actual element values ​​are fitted in the radar frame to construct the actual growth element radar model. The actual growth element radar model is registered and attached to the superimposed base with the predetermined growth element radar model as the superimposed base. After registration and bonding, the established growth element radar model is divided into corresponding sub-killing index radar domains based on the extinction index radar branch. If the actual growth element radar model completely covers the superimposed substrate in the sub-killing index radar domain, the sub-killing index radar domain is marked as a breakthrough domain. The breakthrough division features are preset according to the area of ​​the domain elements that are broken beyond the superimposed substrate when the breakthrough domain is completely covered. If the actual growth element radar model does not completely cover the superimposed substrate in the field of the sub-killing index radar, then the field of the sub-killing index radar is designated as a retention field, and the retention division feature is preset according to the area of ​​the field elements retained inside the superimposed substrate when the retention field is not completely covered. The dosage weight ratio of the pesticide based on the preset growth elements corresponding to the crown of thorns star was obtained by using the application instructions. Based on the dosage weight ratio, a synergistic effect term for the joint dosage consideration between different preset growth elements was constructed. Establish an evaluation root node for eradicating crown-of-thorns starfish. Starting from the evaluation root node, perform synergistic tree structure splitting along the branch paths of each preset growth element using breakthrough and retention splitting features to maximize the synergistic effect term, thereby obtaining the drug application synergistic tree for each preset growth element.

[0013] Furthermore, in a preferred embodiment of the present invention, the step of calculating the leaf node gain contribution during the splitting process of each preset growth element through drug-coordination synergy tree tracking to obtain the actual drug-coordination contribution, accumulating and integrating the actual drug-coordination contributions of each preset growth element to obtain the actual killing dose of the killing agent applied to the target sea area for crown-of-thorns starfish, specifically includes the following steps: The incremental contribution of each preset growth element to the end of all leaf nodes reached by the split in the corresponding drug-dispensing synergy tree is tracked and statistically analyzed, and the change in leaf node gain for each preset growth element is output. The actual contribution of each preset growth element to the application of the pesticide to the crown-of-thorns starfish in the target sea area is determined based on the change in leaf node gain. Construct a metering bubble model for preset growth elements, accumulate the prescribed baseline dose of each preset growth element into the corresponding metering bubble model based on the actual contribution of the drug, obtain the bubble volume value of the metering bubble model, and determine the actual drug ratio of the pesticide for each preset growth element based on the bubble volume value. The dosage of the pesticide was adjusted by combining the actual ratio of all preset growth elements to obtain the actual killing dose of the pesticide applied to the target sea area for crown-of-thorns starfish.

[0014] The beneficial technical effects of this invention are as follows: This invention employs an integrated "identification-positioning-injection" technology approach, forming a complete intelligent operational closed loop. Through a high-definition camera system and deep learning recognition algorithms, it achieves precise detection and real-time positioning of crown-of-thorns starfish, significantly improving target identification accuracy and response speed. The platform module utilizes a multi-electric propulsion layout and a high-precision control system, enabling the equipment to maintain excellent attitude stability and maneuverability even in complex sea conditions, ensuring the injection module can accurately aim at the target in dynamic environments. The injection module uses a PTFE agent storage tank and a motor-driven precision injector structure, automatically adjusting the injection dosage and speed according to the size and posture of the target individual, achieving efficient agent utilization and precise injection. The entire device has automatic depth-keeping, navigation-keeping, and self-balancing functions, allowing for long-term unattended operation. Compared to traditional manual harvesting methods, this equipment significantly improves extermination efficiency and safety, reduces diver risks, lowers labor costs, and effectively prevents secondary pollution to the coral ecosystem caused by agent leakage, demonstrating significant ecological protection and widespread application value. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a three-dimensional structural diagram of the equipment; Figure 2 This is a front view of the structure of the equipment during injection operation. Figure 3 This is a side view of the structure of the device during injection operation.

[0017] The annotations in the attached figures are explained as follows: 101. Platform module; 102. Injection module; 103. Power mechanism; 104. Camera mechanism; 105. Energy mechanism. Detailed Implementation

[0018] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner. Therefore, they only show the components related to the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0019] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the scope of protection of this application. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0021] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0022] like Figures 1-3 As shown, the first aspect of the present invention provides an intelligent extermination device for crown-of-thorns starfish, the device comprising a platform module 101, an injection module 102, and an intelligent identification module.

[0023] The platform module 101 includes a platform frame made of compressed polypropylene material, and the outer shell is made of 6061-T6 hard anodized material. The platform frame is designed to be streamlined to reduce resistance when moving in water. The maximum diving depth is 50m. The platform frame has multiple functional installation areas, and each functional installation area is used to install a sub-functional module.

[0024] The sub-functional modules include a power mechanism 103, a camera mechanism 104, an energy mechanism 105, a control module, an injection module 102, and an intelligent recognition module.

[0025] It should be noted that the platform frame is made of high-strength, corrosion-resistant compressed polypropylene material. This material possesses excellent mechanical properties and corrosion resistance, enabling the equipment to maintain stable operation for extended periods in harsh and complex marine environments. It effectively resists seawater corrosion and high-pressure conditions, ensuring that the structure does not deform or degrade during long-term immersion and repeated operations, thus improving the equipment's service life and reliability. The outer shell is made of 6061-T6 hard anodized aluminum, combining lightweight design with high compressive strength. This reduces the equipment's weight, improves maneuverability, and enhances impact resistance and leak-proof capabilities, providing safety for deep-sea operations. Its streamlined design mimics the shape of certain high-speed swimming organisms in the ocean, effectively reducing water resistance and improving the platform module's mobility.

[0026] The power mechanism 103 is equipped with eight electric thrusters, four of which are horizontally mounted and four are vertically mounted, distributed at different positions on the platform frame to achieve multi-directional thrust output, generating a forward thrust of 24 kgf, a lateral thrust of 14 kgf, and a vertical thrust of 28 kgf. Each electric thruster integrates a propulsion motor, which is a high-torque, low-speed waterproof motor model, and is externally equipped with a waterproof protective shell. The power mechanism 103 is also equipped with a speed sensor and a steering angle sensor.

[0027] It should be noted that the power mechanism 103, through its eight-propeller multi-directional distributed layout, significantly enhances the maneuverability and attitude stability of the crown-of-thorns starfish intelligent extermination device in complex sea conditions. The combination of four horizontal and four vertical propellers enables omnidirectional thrust output, providing flexible control in various movement modes such as forward, backward, lateral, ascent, descent, and attitude adjustment. This allows the device to maintain precise positioning and balance in different currents and depths, ensuring stable execution of extermination operations. Secondly, the propeller motors provide strong torque at low speeds, ensuring the platform has sufficient power for movement and steering in complex marine environments. This provides continuous and stable driving force at low speeds, effectively improving control accuracy during precise positioning, hovering, and low-speed approach to targets. Simultaneously, the waterproof protective shell significantly improves the corrosion and seepage resistance of the waterproof motor, extending the service life of the power mechanism 103 in high-salinity, high-humidity marine environments and reducing maintenance frequency. Furthermore, to achieve precise control over the platform's speed and direction, the power mechanism 103 is equipped with a speed sensor and a steering angle sensor. The speed sensor monitors the platform's speed in real time and feeds the data back to the control module, which then adjusts accordingly based on the set speed value. The steering angle sensor monitors the platform's steering, ensuring it navigates accurately in the predetermined direction and providing crucial data support for precise platform control. The power mechanism 103 improves the equipment's motion accuracy during seabed operations, effectively preventing injection deviations caused by ocean currents or attitude drift, and ensuring accurate and reliable drug injection positioning.

[0028] The camera mechanism 104 is equipped with multiple high-definition PTZ cameras, which are distributed at the front and bottom of the platform frame. Each high-definition PTZ camera is covered with sealing material and protective coating. The camera mechanism 104 is connected to the control module and the intelligent recognition module through a high-speed data transmission line.

[0029] It should be noted that multiple high-definition gimbal cameras are installed at the front and bottom of the platform frame, forming a 360-degree forward and downward field of view coverage. This enables continuous capture of target images under complex seabed terrain and multi-angle attitude changes, achieving real-time tracking and multi-view recognition of crown-of-thorns starfish, avoiding blind spots caused by single-view obstruction. Each high-definition gimbal camera is externally constructed with high-strength sealing materials and protective coatings, effectively preventing seawater infiltration and corrosion, improving the equipment's protective performance and lifespan in high-salinity and high-pressure environments. Simultaneously, the sealing materials and protective coatings reduce image jitter caused by water flow disturbances, ensuring clear and stable images, thereby improving the input quality and accuracy of image recognition algorithms. Internally, the high-definition gimbal cameras are equipped with high-resolution image sensors and high-quality optical lenses, enabling clear imaging of the surrounding environment under various lighting conditions, capturing high-quality images whether on the sunny surface or in the dimly lit deep sea.

[0030] The energy unit 105 integrates a rechargeable battery pack, which is equipped with a power management system.

[0031] It should be noted that the energy mechanism 105, through the integrated design of the rechargeable battery pack and power management system, significantly improves the energy supply stability and endurance of the intelligent starfish eradication device. The rechargeable battery pack uses high-energy-density, long-cycle-life rechargeable battery cells. The high energy density of the battery pack allows it to store a large amount of energy in a small volume and weight. While maintaining a compact size and controllable weight, it can continuously provide stable power to the platform module 101, camera mechanism 104, power mechanism 103, and injection module 102, enabling the device to operate autonomously underwater for extended periods. This avoids operational interruptions caused by frequent power changes or mid-course charging. Furthermore, its excellent low-temperature performance ensures that the battery can operate normally in the cold deep-sea environment. The power management system, as the intelligent energy management center of the energy mechanism 105, can monitor battery charge, voltage, temperature, and charge / discharge status in real time, achieving efficient energy scheduling and safe management through dynamic allocation and protection mechanisms. When the equipment is operating at high power, the energy mechanism 105 can automatically adjust the output power to ensure priority power supply to key modules (such as the power mechanism and intelligent identification module); when the power is insufficient or the temperature is abnormal, the energy mechanism 105 will issue an early warning and take protective measures to effectively prevent safety risks such as battery over-discharge, over-charge and thermal runaway.

[0032] The control module integrates a microprocessor, which connects to and coordinates the control of the power mechanism, camera mechanism 104, energy mechanism 105, and injection module 102. The control module is equipped with a manual mode and an automatic mode. The manual mode is used for equipment debugging and intervention operations in special circumstances, while the automatic mode is used as the main control mode for the equipment during normal operation.

[0033] It should be noted that the microprocessor receives signals from camera arrays, intelligent recognition modules, and various sensors. These signals include information about the surrounding environment, target location, and the device's own status. As the core computing unit, it possesses powerful real-time data processing and multi-task scheduling capabilities, enabling it to simultaneously process visual recognition signals, attitude data, and energy status information in complex marine environments. This ensures coordinated operation of all modules, improving overall response speed and control accuracy. In manual mode, operators can perform equipment debugging, calibration, and intervention operations in special environments. For example, when equipment malfunctions or requires specific testing, operators can manually adjust the equipment for fine-tuning, facilitating troubleshooting and emergency control. In automatic mode, the device autonomously plans its path, controls propulsion, and executes injection based on target information transmitted by the intelligent recognition module, achieving an unmanned, closed-loop extermination process. This significantly improves operational efficiency and accuracy, reducing human interference and errors.

[0034] The injection module 102 includes a drug storage tank, a syringe, an injection needle, a delivery pipe, and connectors. The drug storage tank is made of corrosion-resistant and well-sealed polytetrafluoroethylene material, which can resist the corrosion of different chemical agents. The maximum drug carrying capacity of the drug storage tank is 500ml, and the injection volume at one time is 5ml to 10ml.

[0035] The syringe includes a drug-propelling piston connected to a drive motor, which controls the injection action and stroke of the drug-propelling piston. The delivery tube and connectors are made of medical-grade silicone material, which has good flexibility and chemical stability.

[0036] The intelligent recognition module integrates deep learning algorithms and image recognition technology to perform biological morphological feature recognition on captured images of crown-of-thorns starfish.

[0037] It should be noted that the pesticide storage tank is made of polytetrafluoroethylene (PTFE), a material with extremely high chemical inertness and corrosion resistance. This allows it to withstand various pesticides for extended periods without reaction or leakage, ensuring pesticide purity and storage safety. The well-sealed tank structure effectively prevents pesticide leakage in high-pressure underwater environments. The tank capacity can be designed according to the needs of pest control operations, generally storing enough pesticide for multiple injections, with a maximum capacity of 500ml. This reduces frequent refilling operations and improves operational continuity and efficiency. The syringe is a high-precision, adjustable-dosage model. The precise fit between the internal plunger and barrel allows for accurate control of the injected pesticide volume. A drive motor, following instructions from the control module, precisely adjusts the plunger's displacement, thus strictly controlling the injection volume and speed from 5ml to 10ml per injection. This ensures the optimal dosage for killing starfish of different sizes.

[0038] It should be noted that the injection needle is made of a hard and sharp material, which has the characteristic of strong penetrating power into the body wall of the crown-of-thorns starfish; at the same time, the outer diameter of the injection needle is 2mm, which is not easy to clog and can ensure that the drug can be smoothly injected into the crown-of-thorns starfish through the needle.

[0039] In summary, this device first uses a high-definition pan-tilt camera and image sensor in camera mechanism 104 to perform a comprehensive scan of the surrounding vast target sea area. Then, it uses deep learning algorithms and image recognition technology in the intelligent recognition module to identify all marine life within the target sea area, generating image recognition results. When the recognition results detect the presence of crown-of-thorns starfish, the intelligent recognition module further performs precise pose positioning calculations to determine its real-time location and current attitude. Based on the acquired target location information, platform module 101 controls the power mechanism 103 and control module to start. The propulsion motor of power mechanism 103 drives the electric thruster, which, under the coordinated drive of the control module, flexibly moves the platform module to the area where the target crown-of-thorns starfish is located. Simultaneously, the electric thruster adjusts the device's own attitude according to the current attitude of the crown-of-thorns starfish. At this time, the injection module 102 is activated, precisely aligning the injection needle with the crown-of-thorns starfish. The drive motor drives the injection piston to inject the appropriate amount of the killing agent prepared in the syringe into the crown-of-thorns starfish at the optimal speed and angle, completing the entire killing process.

[0040] The second aspect of this invention provides a control method for an intelligent starfish eradication device, applicable to any of the intelligent starfish eradication devices described in the invention, specifically comprising the following steps: By combining a high-definition gimbal camera and image sensor with an intelligent recognition module, the target sea area is scanned in all directions to obtain underwater image recognition results of crown-of-thorns starfish in the target sea area. The current morphological point cloud of crown-of-thorns starfish in the target sea area is obtained by underwater image recognition results. Taking the crown-of-thorns starfish intelligent extermination device as the field center, the pose of the current morphological point cloud is semantically matched and registered based on the morphological semantic labels of crown-of-thorns starfish under different phenotypic posture conditions to determine the real-time position and current posture of crown-of-thorns starfish. Based on real-time location control, the intelligent starfish eradication device moves to the vicinity of the starfish under the drive of the power mechanism, and controls the injection module to aim at the injection site of the starfish based on the current posture. Obtain the instructions for use of the pesticide, and extract from the instructions for use several preset growth factors of the pesticide against crown of thorns star, the predetermined value of each preset growth factor, and the prescribed baseline dose for different combinations of preset growth factors. Extract the actual element values ​​of crown of thorns star on each preset growth element. Based on the fitting of the predetermined element values ​​and the actual element values, construct a radar model of the predetermined growth element and a radar model of the actual growth element. Based on the breakthrough-stagnation trend of the indicators after the actual growth element radar model is registered with the radar model of the predetermined growth element, perform a collaborative split on the drug application weight ratio of the preset growth element to obtain a drug application collaborative tree for each preset growth element. The leaf node gain contribution during the splitting process of each preset growth element is calculated by tracking the drug application synergy tree to obtain the actual drug application contribution. The actual drug application contribution of each preset growth element is accumulated and integrated to obtain the actual killing dose of the killing agent applied to the target sea area for starfish. The injection module controls the actual killing dose to inject the killing agent into the crown-of-thorns starfish at an appropriate speed and angle.

[0041] Furthermore, in a preferred embodiment of the present invention, the step of obtaining the current morphological point cloud of crown-of-thorns starfish in the target sea area through underwater image recognition results, taking the crown-of-thorns starfish intelligent extermination device as the field center, and performing semantic matching registration transformation on the pose of the current morphological point cloud based on the morphological semantic tags of crown-of-thorns starfish under different phenotypic pose conditions to determine the real-time position and current pose of the crown-of-thorns starfish, specifically includes the following steps: Based on the underwater image recognition results, extract the current morphological point cloud of the crown-of-thorns starfish in the target sea area from the real-time image frame, and at the same time obtain the marine biological knowledge graph based on big data; A semantic segmentation network is constructed, and morphological semantic labels of crown-of-thorns starfish under different phenotypic postures are obtained by retrieving marine biological knowledge graphs. Based on the morphological semantic labels, the current morphological point cloud is formulated in the semantic segmentation network to obtain the morphological semantic terms of each current morphological point cloud. Establish a Manhattan matching matrix, use the Manhattan matching matrix to calculate the Manhattan matching distance of semantic categories between semantic terms of each morphological semantic term, obtain the category matching degree of each point cloud semantic, if the category matching degree is greater than the preset category matching degree, then aggregate the current morphological point cloud corresponding to the semantic annotation term of that morphological term to form a point cloud nearest neighbor matching cluster. Define the current morphological point cloud as the source morphological point cloud, introduce the softmax probability algorithm to calculate the softmax probability distance between the source morphological point cloud and each nearest neighbor point cloud in the nearest neighbor matching cluster, extract only the nearest neighbor point cloud corresponding to the minimum softmax probability distance, label it as the nearest neighbor point cloud, and determine the semantic entropy weight of the nearest neighbor point cloud according to the softmax probability distance, so as to obtain several sets of source morphological point cloud-nearest neighbor point cloud matching point pairs. The LM algorithm is introduced. Based on semantic entropy weight, the residual and Jacobian matrix of each source morphological point cloud-nearest neighbor point cloud matching point pair are calculated in the LM algorithm. The update step size and registration error are obtained respectively. The optimal registration transformation of the current semantic weighted iteration is generated by minimizing the iterative registration error according to the update step size. Based on the current semantic weighted iteration, the optimal registration transformation is used to align and register the source morphological point cloud with the location of the crown-of-thorns starfish intelligent extermination device as the field center. Finally, the pose registration distribution field is output, and the real-time position and current attitude of the crown-of-thorns starfish are determined according to the pose registration distribution field.

[0042] It should be noted that when detecting the traces of crown-of-thorns starfish through image recognition, the starfish undergoes real-time dynamic changes with the flow of seawater. Therefore, traditional attitude positioning methods are difficult to capture the position and movement of the starfish in real time based on the image recognition results. This can easily lead to deviations in the propulsion and injection posture of the extermination equipment, resulting in the incorrect injection of the extermination agent. To address this, our method first acquires the current morphological point cloud of the crown-of-thorns starfish and then uses a semantic segmentation network to assign morphological semantic terms to each point cloud under different phenotypic postures. For example, due to the influence of seawater or its own swimming, the pentagonal body phenotype of the crown-of-thorns starfish may result in a certain arm posture (phenotypic posture) being in a wrist flexion-extension state (morphological semantic term). This yields a set of point clouds with morphological semantic information, ensuring that each point cloud has both a discrete spatial location and a morphological semantic category. This effectively removes noisy point clouds and unifies the coordinate scale for relative registration centered on the device, combining geometric features with semantic understanding. This significantly improves the accuracy of subsequent morphological matching and relative position alignment, especially enhancing the robustness of attitude localization in complex underwater environments (such as occlusion and changes in lighting). Next, the similarity of semantic categories between different morphological semantic terms is measured by Manhattan matching distance. Candidate matches are established only between point clouds with consistent morphological semantics, thereby merging semantically similar morphological point clouds into nearest neighbor matching clusters to form a local aggregation structure. This filters out morphological point clouds that are highly similar in morphological semantics, thereby applying semantic information to the morphological point clouds to further enhance the logical constraints of the matching. This makes the relative attitude matching process of starfish at different locations depend on spatial location on the one hand, and consider the high consistency of semantics that conforms to the morphological behavior of starfish on the other hand, thus improving the stability and convergence speed of real-time attitude matching of starfish and reducing the interference of irrelevant morphological point clouds.

[0043] It should be noted that by using softmax probabilistic distance to quantify the semantic matching confidence between the source point cloud and the nearest neighbor point cloud, high-confidence point cloud matching pairs are obtained through minimum softmax probabilistic distance. Semantically weighted nearest neighbor matching is established using semantic entropy weights to reflect semantic uncertainty, quantifying the reliability of pose localization of matching points and forming a set of semantically weighted matching point pairs. The contribution of the source morphological point cloud-nearest neighbor point cloud matching point pairs is controlled by morphological semantic confidence; higher confidence categories have a greater impact. This effectively transforms traditional geometric nearest neighbor matching into semantic probability space matching, avoiding interference from noisy point clouds and improving the matching accuracy and robustness of real-time pose changes in starfish. Subsequently, the Levenberg-Marquardt (LM) algorithm is used to perform nonlinear least squares optimization on the morphological point cloud registration process. The LM algorithm, also known as the damped least squares algorithm, is a nonlinear least squares optimization algorithm that achieves morphological and semantic weighted nonlinear pose registration iterative optimization. Combined with semantic entropy weights, it performs weighted corrections on matching points of different poses, making the optimization direction smoother and the error smaller. By employing a weighted iterative mechanism for pose registration that integrates semantic credibility, morphological point cloud registration no longer treats all matching points equally but dynamically weighs semantic credibility, thereby improving the accuracy and stability of real-time pose localization and registration for crown-of-thorns starfish. Finally, the iteratively optimized transformation matrix is ​​applied to the source point cloud to align the real-time position of the crown-of-thorns starfish with its current attitude, resulting in a spatially consistent pose registration distribution field. This method integrates semantic matching into relative pose registration, providing more accurate target location and attitude guidance for intelligent extermination equipment in complex marine environments and under conditions of interference from multiple identified individuals. This makes the real-time position and attitude detection of crown-of-thorns starfish more reliable and improves the accuracy of equipment control in dynamic movement and drug injection towards the starfish.

[0044] Furthermore, in a preferred embodiment of the present invention, the step of extracting the actual element values ​​of crown-of-thorns starfish on each preset growth element, constructing a radar model of the preset growth elements and a radar model of the actual growth elements based on the fitting of the predetermined element values ​​and the actual element values, and performing synergistic splitting of the drug application weight ratio of the preset growth elements based on the breakthrough-retention trend of the indicators after the actual growth element radar model is registered with the predetermined growth element radar model to obtain a drug application synergistic tree for each preset growth element, specifically includes the following steps: Create a radar framework, and extend multiple extermination index radar branches dominated by preset growth elements from the center of the radar framework. Fit the predetermined element values ​​to the corresponding extermination index radar branches in the radar framework to generate a radar model of predetermined growth elements for the extermination agent for crown-of-thorns starfish. The actual element values ​​of crown-of-thorns starfish in the target sea area are extracted by underwater image recognition results. The actual element values ​​are fitted in the radar frame to construct the actual growth element radar model. The actual growth element radar model is registered and attached to the superimposed base with the predetermined growth element radar model as the superimposed base. After registration and bonding, the established growth element radar model is divided into corresponding sub-killing index radar domains based on the extinction index radar branch. If the actual growth element radar model completely covers the superimposed substrate in the sub-killing index radar domain, the sub-killing index radar domain is marked as a breakthrough domain. The breakthrough division features are preset according to the area of ​​the domain elements that are broken beyond the superimposed substrate when the breakthrough domain is completely covered. If the actual growth element radar model does not completely cover the superimposed substrate in the field of the sub-killing index radar, then the field of the sub-killing index radar is designated as a retention field, and the retention division feature is preset according to the area of ​​the field elements retained inside the superimposed substrate when the retention field is not completely covered. The dosage weight ratio of the pesticide based on the preset growth elements corresponding to the crown of thorns star was obtained by using the application instructions. Based on the dosage weight ratio, a synergistic effect term for the joint dosage consideration between different preset growth elements was constructed. Establish an evaluation root node for eradicating crown-of-thorns starfish. Starting from the evaluation root node, perform synergistic tree structure splitting along the branch paths of each preset growth element using breakthrough and retention splitting features to maximize the synergistic effect term, thereby obtaining the drug application synergistic tree for each preset growth element.

[0045] It should be noted that precise control of pesticide injection is crucial. However, current control methods typically follow standardized dosage instructions, failing to consider the specific growth characteristics of the crown-of-thorns starfish. This mismatch can result in small doses of pesticide being injected into large starfish, or large doses into small starfish, making it difficult to ensure rapid and complete eradication. Therefore, precise and appropriate dosage control by the injection facility is paramount. To address this, this method constructs a radar model of predetermined growth elements based on the pre-defined values ​​of different growth factors for different insecticides. These predetermined growth factors include body size, health status, activity level, and physiological structure distribution. The predetermined factors are specific parameters of the insecticides used for different dosage ranges targeting the aforementioned predetermined growth factors in crown-of-thorns starfish. For example, the instructions may specify that 2-4 ml of insecticide can only kill crown-of-thorns starfish with a body diameter of 10-45 cm (body size), a tube foot extension ratio of 35-85% (activity level), and 16-22 arms (physiological structure). This is a generalized but ambiguous baseline limitation, making it difficult for traditional methods to accurately determine dosages based on the actual element values ​​of crown-of-thorns starfish. Therefore, an actual growth element radar model of crown-of-thorns starfish is constructed based on the actual element values, which are the specific element parameters of crown-of-thorns starfish in the target sea area. By registering and fitting the actual growth element radar model to the predetermined growth element radar model, a superimposed comparison state of the actual radar indicators and the specified radar indicator range is formed. If the actual growth element radar model completely covers the superimposed substrate (defined growth element radar model) in the sub-infection index radar domain, it indicates that the actual element value is either at the critical point of the defined element value range specified by the pesticide or has exceeded the value range. For example, the actual body diameter of the crown-of-thorns starfish is 48 cm, exceeding the 10-45 cm range specified by the body size element. This is a pre-defined growth element that exceeds the specified value scale, and therefore the corresponding sub-infection index radar domain is marked as an "exceeding domain." Conversely, if the actual element value is not within the defined element value range specified by the pesticide, it indicates that the actual element value is a pre-defined growth element that remains within the specified value scale, and therefore the corresponding sub-infection index radar domain is marked as a "remaining domain."

[0046] It should be noted that the breakthrough and retention features quantify the deviation direction and magnitude of the actual elements of crown-of-thorns starfish from the prescribed element range for pesticide application in each element domain, forming a difference label. Since pesticide dosage often requires proportional adjustment considering the actual situation of each preset growth element, this method establishes a linkage function (synergistic term) between different growth elements by determining the pesticide's weight for each growth element. This describes the mutual reinforcement and offsetting relationships between elements, effectively transforming single-dimensional efficacy control into a multi-dimensional synergistic decision-making problem. It constructs a balanced adjustment mechanism expressing the joint dosage of multiple elements, making pesticide dosage control based on the actual values ​​of multiple preset growth elements more precise. Finally, a tree structure starting from the evaluation root node uses the breakthrough / retention features to drive the splitting and weight update of the synergistic effect term, enabling each element to form an independent dosage adjustment path, thus obtaining multiple dosage synergistic trees, where each tree represents the optimal adjustment strategy for a growth element with respect to the pesticide. This method utilizes radar models to perform coverage analysis on the vector values ​​of actual growth factor parameters within the specified range of the extermination agent. It then uses the breakthrough and retention characteristics derived from the coverage analysis to construct a collaborative tree structure for multi-factor drug formulation decision-making based on the actual conditions of crown-of-thorns starfish. This reduces the drug formulation error rate, minimizes injection failures, ensures complete eradication of crown-of-thorns starfish individuals under different growth conditions, and improves the quality of extermination.

[0047] Furthermore, in a preferred embodiment of the present invention, the step of calculating the leaf node gain contribution during the splitting process of each preset growth element through drug-coordination synergy tree tracking to obtain the actual drug-coordination contribution, accumulating and integrating the actual drug-coordination contributions of each preset growth element to obtain the actual killing dose of the killing agent applied to the target sea area for crown-of-thorns starfish, specifically includes the following steps: The incremental contribution of each preset growth element to the end of all leaf nodes reached by the split in the corresponding drug-dispensing synergy tree is tracked and statistically analyzed, and the change in leaf node gain for each preset growth element is output. The actual contribution of each preset growth element to the application of the pesticide to the crown-of-thorns starfish in the target sea area is determined based on the change in leaf node gain. Construct a metering bubble model for preset growth elements, accumulate the prescribed baseline dose of each preset growth element into the corresponding metering bubble model based on the actual contribution of the drug, obtain the bubble volume value of the metering bubble model, and determine the actual drug ratio of the pesticide for each preset growth element based on the bubble volume value. The dosage of the pesticide was adjusted by combining the actual ratio of all preset growth elements to obtain the actual killing dose of the pesticide applied to the target sea area for crown-of-thorns starfish.

[0048] It should be noted that by tracing down the branches of the drug-coordination synergy tree from the root node to the leaf node, the incremental contribution of the preset growth elements to the overall drug-coordination adjustment is calculated for each split. Finally, the final gain change of all leaf nodes is summarized to measure the response strength of each growth element to the drug ratio adjustment. This yields the gain amplitude of each preset growth element corresponding to the leaf node when considering the actual drug application for crown-of-thorns starfish, i.e., the leaf node gain change. Based on the leaf node gain change, the actual drug application contribution is determined. This actual contribution reflects the weight distribution of the drug's actual effect on the growth element characteristics of the target crown-of-thorns starfish, serving as a scientific basis for subsequent dosage adjustments. This method ensures that the drug application for different crown-of-thorns starfish individuals achieves the killing target while avoiding deviations from the application guidelines, improving the accuracy and appropriateness of drug preparation for killing actual crown-of-thorns starfish by injection institutions, and realizing precise drug injection control for crown-of-thorns starfish.

[0049] The above description, based on preferred embodiments of the present invention, is quite specific and detailed, but it should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A smart extermination device for crown-of-thorns starfish, the device comprising a platform module, an injection module, and a smart identification module, characterized in that: The platform module includes a platform frame made of compressed polypropylene material, and an outer shell made of 6061-T6 hard anodized material. The platform frame is designed to be streamlined to reduce resistance during movement in water. The maximum diving depth is 50m. The platform frame has multiple functional installation areas, and each functional installation area is used to install a sub-functional module. The injection module includes a drug storage tank, a syringe, an injection needle, a delivery tube, and connectors. The drug storage tank is made of corrosion-resistant and well-sealed polytetrafluoroethylene material, which can resist the corrosion of different chemical agents. The maximum drug carrying capacity of the drug storage tank is 500ml, and the injection volume at one time is 5ml to 10ml. The syringe includes a drug-propelling piston connected to a drive motor, which controls the injection action and stroke of the drug-propelling piston. The delivery tube and connectors are made of medical-grade silicone material, which has good flexibility and chemical stability. The intelligent recognition module integrates deep learning algorithms and image recognition technology to perform biological morphological feature recognition on captured images of crown-of-thorns starfish.

2. The intelligent extermination device for crown-of-thorns starfish according to claim 1, characterized in that: The sub-functional modules include a power mechanism, a camera mechanism, an energy mechanism, a control module, an injection module, and an intelligent recognition module.

3. The intelligent extermination device for crown-of-thorns starfish according to claim 2, characterized in that: The power mechanism is equipped with eight electric thrusters, four of which are horizontally mounted and four are vertically mounted, distributed at different positions on the platform frame to achieve multi-directional thrust output, generating a forward thrust of 24 kgf, a lateral thrust of 14 kgf, and a vertical thrust of 28 kgf. Each electric thruster integrates a propulsion motor, which is a high-torque, low-speed waterproof motor model, and is externally equipped with a waterproof protective shell. The power mechanism is also equipped with a speed sensor and a steering angle sensor.

4. The intelligent extermination device for crown-of-thorns starfish according to claim 2, characterized in that: The camera mechanism is equipped with multiple high-definition PTZ cameras, which are distributed at the front and bottom of the platform frame. Each high-definition PTZ camera is covered with sealing material and protective coating. The camera mechanism is connected to the control module and the intelligent recognition module through a high-speed data transmission line.

5. The intelligent extermination device for crown-of-thorns starfish according to claim 2, characterized in that: The energy unit integrates a rechargeable battery pack, which is equipped with a power management system.

6. The intelligent extermination device for crown-of-thorns starfish according to claim 2, characterized in that: The control module integrates a microprocessor, which connects to and coordinates the control of the power mechanism, camera mechanism, energy mechanism, and injection module. The control module is equipped with a manual mode and an automatic mode. The manual mode is used for equipment debugging and intervention operations in special circumstances, while the automatic mode is used as the main control mode for the equipment during normal operation.

7. A control method for an intelligent starfish eradication device, applied to the intelligent starfish eradication device according to any one of claims 1-6, characterized in that, Specifically, the following steps are included: By combining a high-definition gimbal camera and image sensor with an intelligent recognition module, the target sea area is scanned in all directions to obtain underwater image recognition results of crown-of-thorns starfish in the target sea area. The current morphological point cloud of crown-of-thorns starfish in the target sea area is obtained by underwater image recognition results. Taking the crown-of-thorns starfish intelligent extermination device as the field center, the pose of the current morphological point cloud is semantically matched and registered based on the morphological semantic labels of crown-of-thorns starfish under different phenotypic posture conditions to determine the real-time position and current posture of crown-of-thorns starfish. Based on real-time location control, the intelligent starfish eradication device moves to the vicinity of the starfish under the drive of the power mechanism, and controls the injection module to aim at the injection site of the starfish based on the current posture. Obtain the instructions for use of the pesticide, and extract from the instructions for use several preset growth factors of the pesticide against crown of thorns star, the predetermined value of each preset growth factor, and the prescribed baseline dose for different combinations of preset growth factors. Extract the actual element values ​​of crown of thorns star on each preset growth element. Based on the fitting of the predetermined element values ​​and the actual element values, construct a radar model of the predetermined growth element and a radar model of the actual growth element. Based on the breakthrough-stagnation trend of the indicators after the actual growth element radar model is registered with the radar model of the predetermined growth element, perform a collaborative split on the drug application weight ratio of the preset growth element to obtain a drug application collaborative tree for each preset growth element. The leaf node gain contribution during the splitting process of each preset growth element is calculated by tracking the drug application synergy tree to obtain the actual drug application contribution. The actual drug application contribution of each preset growth element is accumulated and integrated to obtain the actual killing dose of the killing agent applied to the target sea area for starfish. The injection module controls the actual killing dose to inject the killing agent into the crown-of-thorns starfish at an appropriate speed and angle.

8. The control method for an intelligent extermination device for crown-of-thorns starfish according to claim 7, characterized in that, The process involves obtaining the current morphological point cloud of crown-of-thorns starfish in the target sea area through underwater image recognition results, using the crown-of-thorns starfish intelligent extermination device as the field center, and performing semantic matching registration transformation on the pose of the current morphological point cloud based on the morphological semantic tags of crown-of-thorns starfish under different phenotypic posture conditions to determine the real-time position and current posture of the crown-of-thorns starfish. Specifically, this includes the following steps: Based on the underwater image recognition results, extract the current morphological point cloud of the crown-of-thorns starfish in the target sea area from the real-time image frame, and at the same time obtain the marine biological knowledge graph based on big data; A semantic segmentation network is constructed, and morphological semantic labels of crown-of-thorns starfish under different phenotypic postures are obtained by retrieving marine biological knowledge graphs. Based on the morphological semantic labels, the current morphological point cloud is formulated in the semantic segmentation network to obtain the morphological semantic terms of each current morphological point cloud. Establish a Manhattan matching matrix, use the Manhattan matching matrix to calculate the Manhattan matching distance of semantic categories between semantic terms of each morphological semantic term, obtain the category matching degree of each point cloud semantic, if the category matching degree is greater than the preset category matching degree, then aggregate the current morphological point cloud corresponding to the semantic annotation term of that morphological term to form a point cloud nearest neighbor matching cluster. Define the current morphological point cloud as the source morphological point cloud, introduce the softmax probability algorithm to calculate the softmax probability distance between the source morphological point cloud and each nearest neighbor point cloud in the nearest neighbor matching cluster, extract only the nearest neighbor point cloud corresponding to the minimum softmax probability distance, label it as the nearest neighbor point cloud, and determine the semantic entropy weight of the nearest neighbor point cloud according to the softmax probability distance, so as to obtain several sets of source morphological point cloud-nearest neighbor point cloud matching point pairs. The LM algorithm is introduced. Based on semantic entropy weight, the residual and Jacobian matrix of each source morphological point cloud-nearest neighbor point cloud matching point pair are calculated in the LM algorithm. The update step size and registration error are obtained respectively. The optimal registration transformation of the current semantic weighted iteration is generated by minimizing the iterative registration error according to the update step size. Based on the current semantic weighted iteration, the optimal registration transformation is used to align and register the source morphological point cloud with the location of the crown-of-thorns starfish intelligent extermination device as the field center. Finally, the pose registration distribution field is output, and the real-time position and current attitude of the crown-of-thorns starfish are determined according to the pose registration distribution field.

9. The control method for an intelligent extermination device for crown-of-thorns starfish according to claim 7, characterized in that, The process involves extracting the actual element values ​​of *Crassula ovata* for each preset growth element, constructing a radar model for the preset growth elements and a radar model for the actual growth elements based on the fitting of the predetermined element values ​​and the actual element values, and then performing a synergistic split on the drug application weight ratio of the preset growth elements based on the breakthrough-retention trend of the indicators after the actual growth element radar model is registered with the predetermined growth element radar model to obtain a drug application synergistic tree for each preset growth element. Specifically, this includes the following steps: Create a radar framework, and extend multiple extermination index radar branches dominated by preset growth elements from the center of the radar framework. Fit the predetermined element values ​​to the corresponding extermination index radar branches in the radar framework to generate a radar model of predetermined growth elements for the extermination agent for crown-of-thorns starfish. The actual element values ​​of crown-of-thorns starfish in the target sea area are extracted by underwater image recognition results. The actual element values ​​are fitted in the radar frame to construct the actual growth element radar model. The actual growth element radar model is registered and attached to the superimposed base with the predetermined growth element radar model as the superimposed base. After registration and bonding, the established growth element radar model is divided into corresponding sub-killing index radar domains based on the extinction index radar branch. If the actual growth element radar model completely covers the superimposed substrate in the sub-killing index radar domain, the sub-killing index radar domain is marked as a breakthrough domain. The breakthrough division features are preset according to the area of ​​the domain elements that are broken beyond the superimposed substrate when the breakthrough domain is completely covered. If the actual growth element radar model does not completely cover the superimposed substrate in the field of the sub-killing index radar, then the field of the sub-killing index radar is designated as a retention field, and the retention division feature is preset according to the area of ​​the field elements retained inside the superimposed substrate when the retention field is not completely covered. The dosage weight ratio of the pesticide based on the preset growth elements corresponding to the crown of thorns star was obtained by using the application instructions. Based on the dosage weight ratio, a synergistic effect term for the joint dosage consideration between different preset growth elements was constructed. Establish an evaluation root node for eradicating crown-of-thorns starfish. Starting from the evaluation root node, perform synergistic tree structure splitting along the branch paths of each preset growth element using breakthrough and retention splitting features to maximize the synergistic effect term, thereby obtaining the drug application synergistic tree for each preset growth element.

10. The control method for an intelligent extermination device for crown-of-thorns starfish according to claim 7, characterized in that, The step of calculating the leaf node gain contribution during the splitting process of each preset growth element through drug-coordination synergy tree tracking to obtain the actual drug-coordination contribution, accumulating and integrating the actual drug-coordination contributions of each preset growth element, and obtaining the actual killing dose of the killing agent applied to the target sea area for crown-of-thorns starfish includes the following steps: The incremental contribution of each preset growth element to the end of all leaf nodes reached by the split in the corresponding drug-dispensing synergy tree is tracked and statistically analyzed, and the change in leaf node gain for each preset growth element is output. The actual contribution of each preset growth element to the application of the pesticide to the crown-of-thorns starfish in the target sea area is determined based on the change in leaf node gain. Construct a metering bubble model for preset growth elements, accumulate the prescribed baseline dose of each preset growth element into the corresponding metering bubble model based on the actual contribution of the drug, obtain the bubble volume value of the metering bubble model, and determine the actual drug ratio of the pesticide for each preset growth element based on the bubble volume value. The dosage of the pesticide was adjusted by combining the actual ratio of all preset growth elements to obtain the actual killing dose of the pesticide applied to the target sea area for crown-of-thorns starfish.

Images