Unmanned aerial vehicle bait precision throwing method and system for gravity-type net cage

Abstract

This application proposes a method and system for precise drone-based bait delivery for gravity-fed net cages, belonging to the field of intelligent aquaculture. The method includes: receiving positioning information from the positioning beacon and configuring a first target area for bait delivery when a bait delivery clock associated with a positioning beacon number is triggered; obtaining the location of an idle drone and performing path optimization based on the first target area for bait delivery to obtain the drone's flight path; inputting the positioning beacon number into a bait delivery scheme library to match a bait delivery scheme; and controlling the idle drone to execute the bait delivery scheme according to the drone's flight path. This application solves the technical problems in the prior art where the movement of floating net cages on the water surface makes it difficult for drones to achieve high-precision positioning and precise bait delivery, achieving the technical effect of precise drone positioning and precise bait delivery for moving net cages through real-time position correction.

Description

Technical Field

[0001] This invention relates to the field of intelligent aquaculture, and in particular to a method and system for precise drone-based bait delivery for gravity-fed net cages. Background Technology

[0002] During fish breeding, proper feeding is crucial for ensuring the stable growth of adult fish. Traditional feeding methods mainly rely on manual labor by boat, which has drawbacks such as high labor intensity, low safety, and low feeding efficiency, making it difficult to meet the needs of large-scale aquaculture.

[0003] With the development of drones, drone technology is gradually being applied to aquaculture. Intelligent feed distribution via drones has become an important research direction for feeding fish fry. Drone feeding offers advantages such as high automation, high efficiency, and low personnel safety risks, effectively addressing the shortcomings of traditional manual feeding methods.

[0004] However, in practical applications, gravity-fed net cages used for fish breeding are affected by environmental factors such as water flow and wind, causing them to continuously shift their position on the water surface. This positional change makes it difficult for drones to accurately locate the net cages, resulting in feed being scattered off-target, making precise feeding impossible, and affecting the normal feeding and growth of adult fish. Summary of the Invention

[0005] This invention addresses the technical problems in the prior art where the movement of gravity-type net cages on the water surface makes it difficult for drones to achieve high-precision positioning and accurate baiting, and provides a method and system for accurate baiting by drones for gravity-type net cages.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: In a first aspect, the present invention provides a method for precise drone baiting in gravity-fed fish cages, comprising: receiving positioning information from a positioning beacon and configuring a first baiting target area when a baiting clock associated with a positioning beacon location is triggered, wherein the positioning beacon is deployed on a gravity-fed fish cage for raising adult fish; obtaining the location of an idle drone and performing path optimization in conjunction with the first baiting target area to obtain a drone flight path; inputting the positioning beacon location into a baiting scheme library and matching a baiting scheme; and controlling the idle drone to execute the baiting scheme according to the drone flight path.

[0007] Secondly, the present invention provides a drone-based precision baiting system for gravity-fed fish cages, comprising: a positioning trigger module, configured to receive positioning information from a positioning beacon and configure a first baiting target area when a baiting clock associated with a positioning beacon number is triggered, wherein the positioning beacon is deployed on the gravity-fed fish cage for raising adult fish; a path optimization module, configured to obtain the location of an idle drone and perform path optimization in conjunction with the first baiting target area to obtain the drone's flight path; a scheme matching module, configured to input the positioning beacon number into a baiting scheme library and match a baiting scheme; and an execution control module, configured to control the idle drone to execute the baiting scheme according to the drone's flight path.

[0008] The beneficial effects of this invention are: When the feeding clock associated with the positioning beacon number is activated, positioning information is received from the positioning beacon. The first target area for feeding is configured. The positioning beacon is deployed on the gravity cages used for raising adult fish, enabling real-time acquisition of the cages' location and providing an accurate target location basis for subsequent precise feeding. The location of the idle drone is obtained, and the drone's flight path is optimized based on the first feeding target area. The optimal flight path is planned according to the relationship between the drone's current location and the target area, improving feeding efficiency. The positioning beacon number is input into the feeding scheme library to match feeding schemes, achieving automated configuration of feeding schemes and ensuring that different cages receive the corresponding feeding parameters. Based on the drone's flight path, the idle drone is controlled to execute the feeding scheme, allowing the drone to fly along the planned path to the target area after real-time position correction and complete the feeding operation according to the matched feeding scheme. This solves the positioning deviation problem caused by changes in the position of the gravity cages.

[0009] Through the above technical solution, the real-time update and precise positioning of the cage's location are achieved based on the positioning beacon deployed on the gravity cage. Combined with path optimization and automatic scheme matching, the drone can accurately throw bait onto the cage, effectively solving the technical problem in the prior art where changes in the position of the gravity cage cause the bait to be unable to be accurately thrown. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the method for precise baiting by drones for gravity-type net cages provided by the present invention. Figure 2 This is a schematic diagram of the structure of the drone-based precision baiting system for gravity-type net cages provided by the present invention.

[0011] In the attached diagram, the components represented by each number are as follows: The module includes a location triggering module 11, a path optimization module 12, a scheme matching module 13, and an execution control module 14. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] In the description of this invention, the terms "first" and "second" 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. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0014] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0015] Example 1, as Figure 1 As shown, this embodiment of the invention provides a method for precise drone baiting for gravity-type net cages, which is applied to the fish breeding control end.

[0016] Specifically, the fry breeding control terminal refers to the system terminal in aquaculture specifically responsible for the management and control of fry breeding. This control terminal undertakes core management functions in the adult fish rearing process, including but not limited to key functions such as feeding scheduling, environmental monitoring, and growth stage tracking. As the intelligent management hub of the entire aquaculture system, the fry breeding control terminal needs to coordinate and handle the following: First, to uniformly manage multiple gravity cages distributed in different waters, and to monitor the growth status and feeding needs of adult fish in each cage in real time; second, to automatically generate and adjust corresponding feeding plans based on factors such as the species, growth stage, and water quality parameters of the adult fish; and third, to allocate available drone resources, optimize flight paths, and ensure that the appropriate amount of feed is accurately delivered to the designated cage area at the right time.

[0017] The fish breeding control terminal receives location information from various positioning beacons and, combined with a preset feeding clock prompt mechanism, can automatically trigger the feeding process, thereby achieving intelligent and automated management of the entire fish breeding process. This effectively solves the technical problems of low efficiency and insufficient accuracy in the traditional manual management mode.

[0018] The method for precise drone bait delivery for gravity-fed cages includes: S1. When the feeding clock associated with the positioning beacon number is triggered, positioning information is received from the positioning beacon, and the first target area for feeding is configured, wherein the positioning beacon is deployed on the gravity cage for raising adult fish.

[0019] Specifically, the feeding process is automatically triggered according to a preset feeding clock mechanism. When the feeding clock associated with a specific positioning beacon number reaches the preset prompt time, the fish breeding control terminal automatically starts the feeding program.

[0020] Positioning beacons, as crucial location sensing devices, are permanently deployed on gravity-fed net cages used for raising adult fish. Each beacon has a unique identifier, known as the beacon tag number, used to distinguish different gravity-fed net cages. When the feeding clock is triggered, the fish breeding control system sends a location query command to the beacon corresponding to its tag number. The beacon then provides its current location information, including but not limited to latitude and longitude coordinates, altitude, and other three-dimensional spatial location data.

[0021] Based on the received positioning information, the fish breeding control terminal performs coordinate transformation and area mapping calculations to determine the specific location of the first target area for bait placement. This configuration process for the first target area for bait placement takes into account the actual size and shape of the gravity-fed net cage and the current positioning information, ensuring that subsequent bait placement accurately covers the effective feeding area within the net cage.

[0022] By using a real-time position correction mechanism based on positioning beacons, the problem of position drift caused by environmental factors such as water flow and waves in gravity cages is effectively solved, providing a reliable positioning basis for subsequent precise feeding.

[0023] S2. Obtain the location of the idle drone, and perform path optimization in conjunction with the target area of ​​the first baiting to obtain the drone's flight path.

[0024] Specifically, in this step, the fish breeding control terminal first sends a status query command to all drones within its jurisdiction via a wireless communication network to obtain the operational status information of each drone. It then filters out drones currently in an idle state and collects their detailed location parameters, including GPS latitude and longitude coordinates, relative sea level altitude, heading angle, and other three-dimensional spatial positioning data. For drones identified as idle, it obtains their remaining battery percentage and selects the drone with the highest remaining battery level from among the idle drones to perform this feeding task. When multiple drones have the same maximum battery level, one of them is randomly selected as the idle drone to perform this feeding task. After identifying the idle drone, its current 3D coordinates are determined using its position parameters. These coordinates serve as the starting point for path planning, with the center coordinates of the identified bait first placement target area as the endpoint, thus initiating 3D path optimization. For example, using an improved A... The pathfinding algorithm comprehensively considers multiple constraints, including minimizing flight distance, energy consumption, and flight time. During path optimization, it first accesses an electronic nautical chart database to obtain terrain information, building distribution, and obstacle data such as power line routes along the flight path, ensuring the planned flight trajectory safely avoids these potential obstacles. Subsequently, based on the UAV's flight performance parameters, including maximum speed, rate of climb, and turning radius, a three-dimensional trajectory conforming to flight dynamics constraints is calculated. Simultaneously, real-time meteorological data, including wind speed, wind direction, and visibility, is acquired to dynamically adjust and optimize the initially planned flight path. Considering the complexity of wind conditions at sea, a relatively stable flight altitude and route are selected to avoid flying in strong wind areas, ensuring the UAV can stably reach the initial bait-dropping target area, thus obtaining the UAV's flight path. The final generated UAV flight path contains a series of three-dimensional waypoint coordinates arranged in chronological order. Each waypoint is marked with detailed flight parameters such as arrival time, flight altitude, flight speed, and heading angle, forming a complete sequence of flight commands. This provides a precise navigation reference and execution basis for the subsequent automated flight control of the idle UAV.

[0025] S3. Input the positioning beacon number into the bait feeding scheme library and match the bait feeding scheme.

[0026] Specifically, the fish breeding control system uses the acquired location beacon number as a query keyword, inputting it into a pre-established feed feeding scheme database for retrieval and matching. This database contains various feeding strategies, storing optimized feeding plans for different adult fish species, growth stages, and environmental conditions. Each plan is associated with a specific location beacon number, ensuring that the corresponding feeding strategy can be quickly located based on the cage's position.

[0027] First, the basic information of the corresponding gravity-fed fish cage is retrieved by locating the beacon number, including key parameters such as the species of adult fish in the cage, their current growth stage, and the size of the cage. Then, a feeding plan matching these parameters is searched in the feeding plan database, including detailed parameter configurations such as feed type, feeding amount, feeding frequency, and feeding time, to obtain the feeding plan.

[0028] The automatic matching mechanism based on the tag number can quickly determine the most suitable feeding scheme for each gravity cage, avoiding the tedious process of manual configuration, while ensuring the adaptability and targeting of the feeding strategy, and providing accurate parameter basis for subsequent precise feeding execution.

[0029] S4. Based on the flight path of the drone, control the idle drone to execute the bait feeding plan.

[0030] Specifically, firstly, the fish breeding control unit fuses the generated drone flight path data with the matched feed delivery scheme parameters to form a complete idle drone. This idle drone includes flight navigation data, distribution control position, and other information. Subsequently, task commands are sent to the idle drone via a wireless communication link, and the idle drone initiates autonomous flight mode upon receiving the task commands. The drone sequentially reaches each waypoint coordinate according to the preset drone flight path, adjusting its attitude, speed, and altitude in real time during flight to ensure that the flight trajectory remains consistent with the drone's flight path.

[0031] When the drone reaches the first target area for bait distribution, its bait delivery device automatically activates, performing the bait distribution action according to the preset feeding mode. During the distribution process, the drone uses onboard sensors to monitor the feeding effect in real time, ensuring that the bait accurately falls into the first target area. After completing the feeding task, the drone autonomously returns to its starting position along the return path and reports the task execution status to the fish breeding control terminal, including information such as feeding completion status, remaining battery power, and equipment status, achieving closed-loop control of the entire feeding process.

[0032] The above technical solution effectively solves the problem of inaccurate feeding positioning caused by the drift of gravity cages, and realizes the precise throwing of bait by drones for gravity cages, improving the accuracy and automation level of bait feeding, while eliminating the safety hazards of manual operation, and providing support for the intelligent development of the aquaculture industry.

[0033] Furthermore, when the bait feeding clock associated with the positioning beacon number is triggered, it includes: S11. When the positioning beacon is deployed on a gravity cage for raising adult fish, the positioning beacon number is associated with the gravity cage number for the adult fish. S12. Obtain the adult fish species, water quality parameters of the breeding water area, and feed nutrition parameters of the adult fish gravity net cage location number; S13. Using the adult fish species, the water quality parameters of the fish breeding area, and the feed nutrition parameters as constraints, retrieve the feed feeding data for the entire cycle in the historical cases of fish breeding. S14. Based on the full-cycle feed feeding data, perform frequent scheme sorting to obtain the selected feed feeding scheme; S15. Based on the selected feed feeding scheme, extract the initial time of the feeding time zone for each adult fish growth stage and configure the prompt time of the feed feeding clock. S16. The selected bait feeding scheme is associated with the positioning beacon number and stored in the bait feeding scheme library.

[0034] In a preferred embodiment, when positioning beacons are deployed on gravity cages used for raising adult fish, a one-to-one correspondence is established between the positioning beacon's tag number and the adult fish's gravity cage tag number. Specifically, each positioning beacon has a unique device code, serving as its tag number, such as "GPS-001," "GPS-002," etc., while each gravity cage also has a corresponding management number, serving as the adult fish's gravity cage tag number, such as "Cage-A01," "Cage-B15," etc. A mapping table is created in the database of the fish breeding control terminal to bind the positioning beacon's tag number to the adult fish's gravity cage tag number one-to-one.

[0035] Subsequently, the information of the adult fish in the gravity-fed net cage was retrieved using the cage's identification number. This included the species of adult fish, water quality parameters of the breeding area, and nutritional parameters of the feed. The species information recorded the specific fish species cultivated in the gravity-fed net cage, such as grass carp, common carp, and silver carp, along with their corresponding strain characteristics. Water quality parameters of the breeding area were collected in real-time by water quality monitoring sensors deployed around the gravity-fed net cage, including water temperature range, pH value, dissolved oxygen content, and ammonia nitrogen concentration. The nutritional parameters of the feed were derived from product specifications provided by the feed supplier, covering information on crude protein content, crude fat content, and vitamin ratios. These information—including the species of adult fish, water quality parameters of the breeding area, and nutritional parameters of the feed—provided accurate constraints for subsequent matching and optimization of feeding programs.

[0036] Next, using the obtained adult fish species, water quality parameters of the breeding waters, and feed nutrient parameters as search constraints, an exact match search was performed on pre-stored historical fish breeding cases to filter out full-cycle feed feeding data that met the current conditions. The historical fish breeding cases consist of a large number of pre-collected and stored successful fish breeding cases, covering complete aquaculture records under different fish species and water environment conditions. The full-cycle feed feeding data includes multiple full-cycle feed feeding schemes, each specifying detailed feeding parameters for each growth stage, including key information such as daily feeding amount, feeding frequency, feed type, and feeding time, as well as corresponding performance indicators such as adult fish survival rate, growth rate, and feed conversion rate. During the search process, a multi-dimensional matching algorithm was used. First, an exact match was performed on the adult fish species to ensure species consistency; then, a range match was performed on the water quality parameters of the breeding waters to filter out cases with similar indicators such as water temperature, pH value, and dissolved oxygen; finally, a similarity calculation was performed on the feed nutrient parameters to select historical records with similar ratios of nutrients such as protein and fat. Through a multi-layered screening mechanism, multiple full-cycle feeding schemes that best match the current conditions can be accurately extracted from historical cases of fish breeding. These schemes serve as full-cycle feeding data, providing high-quality candidate data for subsequent frequent scheme selection.

[0037] Then, the multiple full-cycle feeding schemes in the acquired full-cycle feeding data were subjected to frequent scheme selection. By calculating the frequency of each full-cycle feeding scheme in historical cases, the most commonly used and effective feeding patterns were identified, and the selected feeding scheme was obtained. Specifically, each full-cycle feeding scheme was first decomposed according to the adult fish growth stages, including the fry stage, larval stage, baby fry stage, summer fry stage, juvenile stage, young fish stage, and adult stage. The feeding parameters for each stage constituted a feeding sub-scheme. Next, frequent itemset mining was used to count the frequency of various feeding parameter combinations in multiple full-cycle feeding schemes, and a weighted score was calculated based on the corresponding performance indicators, including key performance indicators such as adult fish survival rate, average daily weight gain, and feed conversion rate. Finally, the feeding pattern combination with the highest frequency and the best overall effect was selected as the chosen feeding scheme.

[0038] Next, based on the selected feeding plan, the initial time of the feeding time zone for each adult fish growth stage is extracted to configure the prompt time for the feeding clock. The feeding time arrangement for each growth stage in the selected feeding plan is analyzed. For example, during the fry stage, feeding is typically set to 4 times per day, with a feeding time zone of 6:00-18:00 and an initial time of 6:00; during the larval stage, feeding is set to 3 times per day, with a feeding time zone of 7:00-17:00 and an initial time of 7:00, etc. The initial time of the feeding time zone for each growth stage is set as the prompt time for the feeding clock to ensure that the feeding process is automatically triggered at the appropriate time.

[0039] Next, the selected bait feeding plan is associated with the location beacon number and stored in the bait feeding plan library. By establishing the association between the selected bait feeding plan and the corresponding location beacon number, personalized feeding strategy data is formed. The bait feeding plan library provides accurate parameter configuration and time scheduling information for subsequent automated feeding execution.

[0040] Through the above steps, an intelligent bait feeding scheme based on the positioning beacon number was generated and configured, which effectively solved the problems of cumbersome and subjective nature of traditional manual configuration of feeding schemes. The data-driven approach ensured the reliability and pertinence of the feeding strategy, and laid the foundation for subsequent precise drone feeding.

[0041] Furthermore, based on the full-cycle feed feeding data, frequent scheme sorting is performed to obtain selected feed feeding schemes, including: S141. Perform pairwise scheme deviation evaluation on the full-cycle feed feeding data to obtain a set of feed feeding scheme deviation parameters; S142. Based on the set of deviation parameters of the bait feeding scheme, using the deviation parameters of the bait feeding scheme as distance measurement parameters, perform LOF outlier calculation for each full-cycle bait feeding scheme to obtain the set of outlier factors of the bait feeding scheme. S143. Extract the full-cycle feeding scheme corresponding to the minimum value of the outlier set of the feeding scheme, and set it as the selected feeding scheme.

[0042] In a preferred embodiment, pairwise bias assessments are performed on the full-cycle feeding data to obtain a set of feeding scheme bias parameters. Specifically, multiple full-cycle feeding schemes are extracted from the full-cycle feeding data, and then pairwise comparisons are performed on these schemes. For any two full-cycle feeding schemes, the differences in feeding parameters at each adult fish growth stage are calculated, including the deviation values ​​of parameters such as feeding amount, feeding frequency, and feed type. The normalized values ​​of the deviations of parameters with the same attribute are calculated as the dimensional bias, and Euclidean distance is used for comprehensive calculation to obtain the feeding scheme bias parameters between the two full-cycle feeding schemes. This process is repeated until all pairwise comparisons of the schemes are completed, ultimately forming a set of feeding scheme bias parameters.

[0043] Next, based on the set of feed feeding scheme deviation parameters, and using these parameters as a distance metric, the LOF (Location of Observation) outlier factor for each full-cycle feed feeding scheme is calculated, resulting in a set of outlier factors. The LOF outlier factor is an algorithm used to detect the degree of anomalies in data points. Each full-cycle feed feeding scheme is treated as a data point, and the local density deviation of each scheme relative to its neighboring schemes is calculated using the feed feeding scheme deviation parameters as a distance metric. Specifically, firstly, for each full-cycle feed feeding scheme, the k nearest neighbor schemes are found from all full-cycle feed feeding schemes based on the feed feeding scheme deviation parameters; then, the average distance from this scheme to its k neighbor schemes is calculated, called the reachability distance; next, the local reachability density of this scheme is calculated, which is the reciprocal of the reachability distance of its k neighbor schemes; finally, the local reachability density of this scheme is compared with the local reachability densities of its k neighbor schemes to obtain the LOF value. The LOF value is calculated as (sum of local reachable densities of k neighboring schemes / k) / local reachable density of the scheme. A smaller LOF value indicates that the full-cycle feeding scheme is more similar to its neighboring schemes, i.e., it conforms more to the mainstream feeding pattern; a larger LOF value indicates that the scheme is more unusual or special compared to its neighboring schemes.

[0044] Subsequently, the full-cycle feeding scheme corresponding to the minimum value of the outlier set of the feeding schemes is extracted and set as the selected feeding scheme. All LOF outlier values ​​in the outlier set of the feeding schemes are traversed to find the minimum value. The full-cycle feeding scheme corresponding to this minimum value is the most similar and representative mainstream feeding strategy to other schemes. By selecting the scheme with the smallest LOF value as the selected feeding scheme, the stability and reliability of the selected scheme are ensured, avoiding feeding risks caused by using abnormal or extreme schemes.

[0045] Furthermore, pairwise scheme deviation evaluation is performed on the full-cycle feed feeding data to obtain a set of feed feeding scheme deviation parameters, including: S1411. Extract a first full-cycle feeding scheme from the full-cycle feeding data, wherein the first full-cycle feeding scheme has a first full-cycle first stage feeding scheme up to the first full-cycle Nth stage feeding scheme. S1412. Extract a second full-cycle feeding scheme from the full-cycle feeding data, wherein the second full-cycle feeding scheme includes a second full-cycle first-stage feeding scheme up to a second full-cycle Nth-stage feeding scheme. S1413. Using the normalized value of the deviation of the same attribute parameters of the first full-cycle first-stage feed feeding scheme and the second full-cycle first-stage feed feeding scheme as the dimensional deviation, calculate the Euclidean distance to obtain the first-stage Euclidean distance. S1414. Until the normalized value of the deviation of the same attribute parameter between the first full-cycle Nth stage feeding scheme and the second full-cycle Nth stage feeding scheme is used as the dimensional deviation, calculate the Euclidean distance to obtain the Nth stage Euclidean distance. S1415. Obtain the weights from the first stage to the Nth stage based on the Delphi method weighting, perform a weighted summation of the Euclidean distance from the first stage to the Nth stage to obtain the first feed feeding scheme deviation parameter, and add it to the set of feed feeding scheme deviation parameters.

[0046] In a preferred embodiment, firstly, an arbitrary full-cycle feeding scheme is extracted from the full-cycle feeding data as the first full-cycle feeding scheme. The first full-cycle feeding scheme covers the complete growth cycle of fish fry from hatching to adulthood, specifically including the first stage feeding scheme up to the Nth stage feeding scheme. Here, N represents the total number of stages in adult fish growth, typically including seven stages: fry, larvae, juveniles, summer fry, young fish, and adult fish. Each stage's feeding scheme includes corresponding parameters such as feeding amount, feeding frequency, and feed type. Then, another full-cycle feeding scheme is extracted from the full-cycle feeding data as the second full-cycle feeding scheme. The second full-cycle feeding scheme also covers the complete adult fish growth cycle, including the first stage feeding scheme up to the Nth stage feeding scheme, and its structure corresponds to the first full-cycle feeding scheme, facilitating stage-by-stage parameter comparison.

[0047] Subsequently, the Euclidean distance between corresponding stages is calculated. Taking the first stage feeding scheme of the first full cycle and the first stage feeding scheme of the second full cycle as examples, parameters with the same attributes in the two schemes, such as feeding amount, feeding frequency, and feed particle size, are extracted. The deviation values ​​of each parameter are calculated and normalized to obtain the normalized deviation values ​​of each dimension. Then, these normalized deviation values ​​are used as dimensional deviations, and the Euclidean distance formula is used to calculate the distance between the two stage schemes to obtain the Euclidean distance of the first stage. The above calculation process is repeated to calculate the Euclidean distance from the second stage to the Nth stage, until the Euclidean distance of the Nth stage is obtained. By calculating the distance stage by stage, the parameter differences between the two full-cycle feeding schemes at each growth stage can be comprehensively reflected.

[0048] Subsequently, based on expert experience and the growth patterns of adult fish, the Delphi method was used to assign weights for the first stage up to the Nth stage. These weights reflect the importance of different growth stages to the overall feeding effect. Specifically, the Delphi method is a structured expert consultation method that invites multiple professionals with extensive aquaculture experience, including fish nutrition experts, aquaculture engineers, and feed formulators, to score the importance of different growth stages based on their respective expertise. The experts comprehensively considered key indicators such as the physiological characteristics of adult fish at each stage, changes in nutritional requirements, and factors affecting survival rates. Through multiple rounds of anonymous scoring and feedback correction, a consensus was reached to obtain the weights for the first stage up to the Nth stage. Then, the weighted summation of the Euclidean distances from the first stage to the Nth stage was performed to calculate the first-stage feeding scheme deviation parameter between the first and second-stage full-cycle feeding schemes. This parameter was added to the feeding scheme deviation parameter set as a comprehensive measure of the difference between the two schemes.

[0049] Then, the above deviation calculation process is repeated for other full-cycle feeding schemes in the full-cycle feeding data. Specifically, the first full-cycle feeding scheme is compared pairwise with each of the remaining full-cycle feeding schemes, and the corresponding feeding scheme deviation parameters are obtained using the same calculation method. Simultaneously, other full-cycle feeding schemes are compared pairwise to ensure that any two schemes in the full-cycle feeding data can obtain corresponding feeding scheme deviation parameters. By traversing all possible scheme combinations, a complete set of feeding scheme deviation parameters is finally obtained. This set of feeding scheme deviation parameters contains a comprehensive difference metric between any two schemes in the full-cycle feeding data. This comprehensive pairwise comparison mechanism provides a complete distance metric basis for subsequent LOF outlier factor calculation, ensuring that each full-cycle feeding scheme can accurately assess its similarity and representativeness relative to other schemes.

[0050] Furthermore, receiving positioning information from the positioning beacon and configuring the first target area for bait placement includes: S17. Obtain the three-dimensional mesh model of the gravity cage; S18. Extract the relative coordinate parameters of the bait receiving surface mesh model and the positioning beacon deployment position in the three-dimensional mesh model of the gravity cage, wherein the bait receiving surface mesh model is the opening area on the water surface of the cage, which can receive bait being thrown. S19. When receiving positioning information from the positioning beacon, the coordinate position of the bait receiving surface grid model is located using the positioning information as the origin and the relative coordinate parameters. Then, a world coordinate transformation is performed to obtain the first bait throwing target area.

[0051] In a preferred embodiment, firstly, a three-dimensional mesh model of the gravity-type cage is obtained. The three-dimensional mesh model of the gravity-type cage is a digital three-dimensional representation of the actual gravity-type cage, constructed through measurement and modeling techniques. This three-dimensional mesh model accurately describes the geometric structural features of the gravity-type cage, including its length, width, and height dimensions, as well as the spatial relationships of its various components such as the cage frame, mesh panels, and floats. The three-dimensional mesh model of the gravity-type cage uses a standard three-dimensional coordinate system, providing an accurate geometric reference for subsequent spatial positioning calculations.

[0052] Subsequently, the relative coordinate parameters of the feed receiving surface mesh model and the positioning beacon deployment location were extracted from the 3D mesh model of the gravity-fed net cage. The feed receiving surface mesh model corresponds to the opening area on the water surface of the net cage, i.e., the effective area where adult fish can receive feed. The boundary contour of this opening area was identified in the 3D mesh model of the gravity-fed net cage, and the spatial offset of its geometric center point or key feature point relative to the positioning beacon deployment location was calculated, including the relative distance values ​​in the X, Y, and Z axis directions. These relative coordinate parameters describe the fixed spatial relationship between the feed receiving surface and the positioning beacon.

[0053] Upon receiving positioning information from the positioning beacon, the system uses this information as the origin of the spatial coordinate system and combines it with the acquired relative coordinate parameters to determine the absolute spatial position of the bait receiving surface grid model at the current moment through coordinate calculations. Subsequently, a world coordinate transformation is performed to convert the position information in the local coordinate system into a unified global coordinate system representation, ultimately determining the precise geographic coordinate range of the first bait placement target area, providing accurate target positioning data for precise drone feeding.

[0054] Furthermore, based on the flight path of the drone, the idle drone is controlled to execute the bait feeding scheme, including: S41. Collect the wind speed and wind direction parameters of the first baiting target area; S42. Several bait throwing positions are evenly distributed at preset distances within a preset range above the first bait throwing target area; S43. Extract the feed density, feed particle size and feeding amount from the feed feeding scheme; S44. By using the feed waste prediction network, traversing the several feed scattering locations, and combining the first feed scattering target area, the wind speed parameter, the wind direction parameter, the feed density, the feed particle size, and the feeding amount to perform analysis and obtain several feed waste prediction values. S45. Based on the feeding amount, iterate through the several predicted values ​​of feed waste and calculate the several effective feeding amounts; S46. When the effective feeding amount of the plurality of baits is greater than or equal to the effective feeding amount threshold, the corresponding bait scattering position is randomly selected and the bait feeding scheme is executed. S47. When none of the predicted values ​​of feed waste amount are greater than or equal to the effective feeding amount threshold, select the feed throwing position with the minimum value among the predicted values ​​of feed waste amount. S48. Sum the minimum value of the predicted feeding amount and the predicted amount of feed waste to obtain an updated feed feeding plan, and execute the updated feed feeding plan at the minimum feed scattering position.

[0055] In a preferred embodiment, firstly, wind speed and direction parameters are collected in the target area for the first bait placement. Specifically, current environmental wind conditions are acquired in real time using meteorological sensors deployed around the target area. Wind speed parameters record the intensity of the wind, typically measured in meters per second; wind direction parameters record the direction and angle of the wind, measured with true north as the 0° reference. Both wind speed and direction parameters have a significant impact on the flight trajectory and final landing point of the bait.

[0056] Then, within a preset range above the initial baiting target area, several baiting positions are evenly distributed at preset distances. The preset range is a three-dimensional spatial area centered on the initial baiting target area, extending upwards to a certain height. Its boundaries are set by the user based on actual feeding needs and the drone's flight performance. The preset distance is a spatial interval value determined based on the drone's feeding accuracy requirements and the target area size, and can also be adjusted by the user according to specific feeding accuracy requirements. Within the preset range, multiple candidate baiting positions are evenly distributed in a grid pattern at preset distances, forming a three-dimensional array of feeding points covering the target area, resulting in several baiting positions.

[0057] Subsequently, key bait characteristic parameters were extracted from the bait feeding program, including bait density, bait particle size, and feeding amount. Bait density reflects the physical compactness of the bait and affects its falling speed in the air; bait particle size indicates the size of the bait particles and affects its susceptibility to wind; feeding amount specifies the total weight of bait for this feeding.

[0058] Then, a bait waste prediction network is used to analyze and evaluate each baiting location among several baiting positions. This bait waste prediction network is an intelligent prediction model trained based on machine learning algorithms, which can comprehensively consider multiple factors such as bait characteristics, environmental conditions, and baiting location. The network iterates through several baiting locations, inputting the coordinates of each location along with input parameters such as the first baiting target area, wind speed parameters, wind direction parameters, bait density, bait particle size, and feeding amount into the bait waste prediction network. The network calculates the corresponding predicted bait waste value, resulting in several predicted bait waste values. Next, based on the feeding amount and the several predicted bait waste values, several effective feeding amounts are calculated. For each baiting location, the same feeding amount is used for calculation. The effective feeding amount for that location is obtained by subtracting the predicted bait waste value from this feeding amount; that is, the weight of bait that would successfully fall into the first baiting target area and be ingested by adult fish if bait were placed at that location.

[0059] When an effective feeding amount is greater than or equal to the effective feeding amount threshold among several effective feeding amounts, it indicates that a feeding location can meet the preset feeding effect requirements. At this point, a location is randomly selected from those locations as the final feeding location, and the original feeding plan is executed. Since the effective feeding amount at these locations has already reached the threshold requirement, no adjustment to the feeding amount is needed; simply executing the original plan will achieve the expected feeding effect.

[0060] When the effective feeding amounts for several baiting locations are all less than the effective feeding amount threshold, it indicates that none of the baiting locations can achieve the ideal feeding effect under the current environmental conditions, posing a significant risk of bait waste. In this case, the smallest predicted bait waste value is selected from several predicted values, and this minimum predicted bait waste value is determined as the optimal feeding location, i.e., the minimum waste feeding location. Simultaneously, for the selected minimum waste feeding location, the expected bait loss is compensated by increasing the feeding amount. Specifically, the original feeding amount is added to the minimum predicted bait waste value for that location to obtain an updated feeding plan. This compensation mechanism ensures that when feeding is performed at the minimum waste feeding location, the effective feeding amount, after deducting the expected waste, can meet the nutritional needs of the adult fish, achieving the goal of precise feeding.

[0061] Through the above steps, wind speed and direction parameters are collected in real time and combined with a bait waste prediction network to accurately assess the feeding effect at different throwing locations, effectively addressing the impact of environmental factors on bait throwing accuracy. Multiple candidate location configurations and intelligent location selection strategies ensure high-precision feeding even under complex weather conditions. When ideal feeding conditions are not met, a compensation mechanism that dynamically adjusts the feeding amount ensures that the effective feeding amount reaches the expected target, avoiding insufficient feeding due to environmental factors. This improves the environmental adaptability and feeding accuracy of drone bait feeding, providing a reliable technical guarantee for achieving automated feeding.

[0062] Furthermore, by using a feed waste prediction network, traversing the various feed placement locations, and combining the first feed placement target area, the wind speed parameter, the wind direction parameter, the feed density, the feed particle size, and the feeding amount, analysis is performed to obtain several predicted feed waste values, including: S441. Collect multiple sets of historical data on bait placement, wherein any set of the multiple sets of historical data on bait placement includes bait placement location record data, bait placement target area record data, wind speed record data, wind direction record data, bait density record data, bait particle size record data, feeding amount record data, and bait waste amount record data. S442. Using the recorded data of the amount of feed waste as supervision, and the recorded data of the feed throwing location, the recorded data of the feed throwing target area, the recorded data of the wind speed, the recorded data of the wind direction, the recorded data of the feed density, the recorded data of the feed particle size, and the recorded data of the feeding amount as input, the feed waste prediction network is trained using machine learning.

[0063] In a preferred embodiment, multiple sets of historical bait distribution data are collected as the base dataset for network training. These sets of historical bait distribution data originate from previous drone feeding operation records, with each set constituting a complete feeding case sample. Each set of historical bait distribution data includes records of bait distribution location, target area, wind speed, wind direction, bait density, bait particle size, feeding amount, and bait waste. Among them, the bait placement location data records the three-dimensional coordinate position of the drone when it performs the feeding; the bait placement target area data records the coordinate range and geometric characteristics of the target cage area; the wind speed and wind direction data record the environmental wind conditions at the time of feeding, such as wind speed of 3.2 m / s and wind direction of 30° north of east; the bait density data, bait particle size data, and feeding amount data record the physical characteristics and weight of the bait used; and the bait waste data record the weight of the bait that failed to fall into the target area during the feeding.

[0064] Subsequently, a supervised learning method was used to train a feed waste prediction network. During training, the recorded feed waste data served as the supervision label, representing the network's expected output target; the recorded data for feed placement location, target area, wind speed, wind direction, feed density, feed particle size, and feed amount served as the input feature vector. Machine learning algorithms, such as neural networks, support vector machines, or random forests, were used to train the network to learn the nonlinear mapping relationship between the input features and feed waste. After multiple rounds of iterative training and parameter optimization, a feed waste prediction network capable of accurately predicting feed waste under different conditions was finally obtained, providing intelligent decision support for location selection and feed amount adjustment in actual feeding operations. Taking a neural network as an example, a multi-layer feedforward neural network was constructed, including an input layer, hidden layers, and an output layer. The input layer had seven neurons, receiving input parameters such as the three-dimensional coordinates of the placement location, the geometric features of the target area, wind speed, wind direction, feed density, feed particle size, and feed amount. The hidden layer employs a two-layer structure: the first hidden layer contains 64 neurons, and the second hidden layer contains 32 neurons, both using the ReLU activation function to enhance the network's non-linear fitting ability. The output layer contains one neuron, directly outputting the predicted amount of bait waste. During training, the input features of historical data are normalized before being input into the network. The predicted value is calculated through forward propagation, and then the mean squared error loss function is calculated by comparing it with the actual bait waste labels. The Adam optimization algorithm is used for backpropagation to update the network weights. After iterative training, an accurate prediction model between the input features and the amount of bait waste can be established, resulting in a bait waste prediction network.

[0065] By constructing a feed waste prediction network, it is possible to predict feed waste under different environmental conditions and at different distribution locations. When faced with complex and ever-changing feeding environments, the prediction network can quickly assess the impact of various combinations of factors on feeding effectiveness, providing a basis for the drone to select the optimal distribution location. By comparing and analyzing the predicted waste values ​​of multiple candidate locations, feed utilization efficiency can be maximized while meeting feeding requirements. Furthermore, when the feeding effect at all candidate locations is unsatisfactory, it can provide a precise basis for the dynamic adjustment of the feeding amount, ensuring that the expected effective feeding amount can be achieved even under adverse environmental conditions, thereby improving the environmental adaptability and feeding accuracy of drone feed delivery.

[0066] Example 2, as Figure 2 As shown, based on the same inventive concept as the drone baiting precision throwing method for gravity-type net cages provided in Embodiment 1, this embodiment of the invention also provides a drone baiting precision throwing system for gravity-type net cages, comprising: The positioning trigger module 11 is used to receive positioning information from the positioning beacon and configure the first target area for baiting when the bait feeding clock associated with the positioning beacon number is triggered. The positioning beacon is deployed on the gravity net cage for raising adult fish. The path optimization module 12 is used to obtain the location of the idle drone and perform path optimization in conjunction with the target area of ​​the first bait scattering to obtain the drone's flight path; The scheme matching module 13 is used to input the positioning beacon number into the bait feeding scheme library and match the bait feeding scheme. The execution control module 14 is used to control the drone to execute the bait feeding plan according to the drone's flight path.

[0067] Furthermore, the execution steps of the positioning trigger module 11 include: When the positioning beacon is deployed on a gravity cage for raising adult fish, the positioning beacon number is associated with the gravity cage number for the adult fish. Obtain the adult fish species, water quality parameters of the breeding area, and feed nutrition parameters of the adult fish in the gravity net cage. Using the adult fish species, the water quality parameters of the fish breeding area, and the feed nutrition parameters as constraints, the feed feeding data for the entire cycle is retrieved from the historical cases of fish breeding. Based on the full-cycle feed feeding data, frequent scheme sorting is performed to obtain the selected feed feeding scheme; Based on the selected feed feeding scheme, the initial time of the feeding time zone for each adult fish growth stage is extracted and the prompt time of the feed feeding clock is configured. The selected bait feeding scheme is associated with the location beacon number and stored in the bait feeding scheme library.

[0068] Furthermore, the execution steps of the positioning trigger module 11 also include: The pairwise scheme deviation evaluation was performed on the full-cycle feed feeding data to obtain a set of feed feeding scheme deviation parameters; Based on the set of deviation parameters of the feeding scheme, and using the deviation parameters of the feeding scheme as distance metric parameters, the LOF outlier factor of each full-cycle feeding scheme is calculated to obtain the set of outlier factors of the feeding scheme. Extract the full-cycle feeding scheme corresponding to the minimum value of the outlier set of the feeding scheme, and set it as the selected feeding scheme.

[0069] Furthermore, the execution steps of the positioning trigger module 11 also include: From the full-cycle feeding data, a first full-cycle feeding scheme is extracted, wherein the first full-cycle feeding scheme includes a first full-cycle first-stage feeding scheme up to a first full-cycle Nth-stage feeding scheme. From the full-cycle feeding data, a second full-cycle feeding scheme is extracted, wherein the second full-cycle feeding scheme includes the second full-cycle first-stage feeding scheme up to the second full-cycle Nth-stage feeding scheme. Using the normalized value of the deviation of the same attribute parameters of the first full-cycle first-stage feeding scheme and the second full-cycle first-stage feeding scheme as the dimensional deviation, the Euclidean distance is calculated to obtain the first-stage Euclidean distance. Until the normalized value of the deviation of the same attribute parameter between the first full-cycle Nth stage feeding scheme and the second full-cycle Nth stage feeding scheme is used as the dimensional deviation, the Euclidean distance is calculated to obtain the Nth stage Euclidean distance. Obtain the first stage weights up to the Nth stage weights based on the Delphi method, perform a weighted summation of the Euclidean distances from the first stage to the Nth stage, obtain the first feed feeding scheme deviation parameter, and add it to the set of feed feeding scheme deviation parameters.

[0070] Furthermore, the execution steps of the positioning trigger module 11 also include: Obtain a 3D mesh model of a gravity-type wire cage; Extract the relative coordinate parameters of the bait receiving surface mesh model and the positioning beacon deployment position in the three-dimensional mesh model of the gravity cage. The bait receiving surface mesh model is the opening area on the water surface of the cage, which can receive bait. When receiving positioning information from the positioning beacon, the coordinate position of the bait receiving surface grid model is located using the positioning information as the origin and the relative coordinate parameters. Then, a world coordinate transformation is performed to obtain the first bait throwing target area.

[0071] Furthermore, the execution steps of the execution control module 14 include: Collect wind speed and wind direction parameters in the target area where the bait is first scattered; Several bait throwing positions are evenly distributed at preset distances within a preset range above the first bait throwing target area; From the aforementioned feeding scheme, extract the feed density, feed particle size, and feeding amount; By using a feed waste prediction network, the network iterates through several feed placement locations and performs analysis based on the first feed placement target area, the wind speed parameter, the wind direction parameter, the feed density, the feed particle size, and the feeding amount to obtain several predicted feed waste values. Based on the feeding amount, the effective feeding amount is calculated by iterating through the predicted values ​​of several feed waste amounts; When the effective feeding amount of the aforementioned baits is greater than or equal to the effective feeding amount threshold, the corresponding bait scattering position is randomly selected, and the bait feeding scheme is executed. When none of the predicted values ​​for feed waste are greater than or equal to the effective feeding threshold, the feed scattering position is selected based on the minimum value among the predicted values ​​for feed waste. The minimum value of the sum of the feeding amount and the predicted amount of feed waste is used to obtain an updated feed feeding plan, which is then executed at the minimum feed scattering location.

[0072] Furthermore, the execution steps of the execution control module 14 also include: Collect multiple sets of historical data on bait placement. Each set of these historical data includes data on bait placement location, bait placement target area, wind speed, wind direction, bait density, bait particle size, feeding amount, and bait waste. Using the recorded data on feed waste as supervision, and the recorded data on feed placement location, feed placement target area, wind speed, wind direction, feed density, feed particle size, and feed amount as input, a feed waste prediction network is trained using machine learning.

[0073] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0074] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0076] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0077] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0078] Although preferred embodiments of the invention have been described, those skilled in the art, once they have learned the basic inventive concept, can make other changes and modifications to these embodiments.

[0079] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of this invention and its equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for precise bait throwing by unmanned aerial vehicle for gravity type net pens, characterized in that, Applications include: When the feeding clock associated with the positioning beacon number is triggered, positioning information is received from the positioning beacon, and the first target area for feeding is configured, wherein the positioning beacon is deployed on the gravity cage for raising adult fish; The location of the idle drone is obtained, and path optimization is performed in conjunction with the target area of ​​the first bait placement to obtain the drone's flight path; Input the location beacon number into the bait feeding scheme library to match the bait feeding scheme; Based on the flight path of the drone, control the idle drone to execute the bait feeding plan.

2. The method as described in claim 1, characterized in that, When the bait feeding clock associated with the location beacon number is triggered, including: When the positioning beacon is deployed on a gravity cage for raising adult fish, the positioning beacon number is associated with the gravity cage number for the adult fish. Obtain the adult fish species, water quality parameters of the breeding area, and feed nutrition parameters of the adult fish in the gravity net cage. Using the adult fish species, the water quality parameters of the fish breeding area, and the feed nutrition parameters as constraints, the feed feeding data for the entire cycle is retrieved from the historical cases of fish breeding. Based on the full-cycle feed feeding data, frequent scheme sorting is performed to obtain the selected feed feeding scheme; Based on the selected feed feeding scheme, the initial time of the feeding time zone for each adult fish growth stage is extracted and the prompt time of the feed feeding clock is configured. The selected bait feeding scheme is associated with the location beacon number and stored in the bait feeding scheme library.

3. The method as described in claim 2, characterized in that, Based on the full-cycle feed feeding data, frequent scheme sorting is performed to obtain selected feed feeding schemes, including: The pairwise scheme deviation evaluation was performed on the full-cycle feed feeding data to obtain a set of feed feeding scheme deviation parameters; Based on the set of deviation parameters of the feeding scheme, and using the deviation parameters of the feeding scheme as distance metric parameters, the LOF outlier factor of each full-cycle feeding scheme is calculated to obtain the set of outlier factors of the feeding scheme. Extract the full-cycle feeding scheme corresponding to the minimum value of the outlier set of the feeding scheme, and set it as the selected feeding scheme.

4. The method as described in claim 3, characterized in that, The pairwise scheme deviation evaluation was performed on the full-cycle feed feeding data to obtain a set of feed feeding scheme deviation parameters, including: From the full-cycle feeding data, a first full-cycle feeding scheme is extracted, wherein the first full-cycle feeding scheme includes a first full-cycle first-stage feeding scheme up to a first full-cycle Nth-stage feeding scheme. From the full-cycle feeding data, a second full-cycle feeding scheme is extracted, wherein the second full-cycle feeding scheme includes the second full-cycle first-stage feeding scheme up to the second full-cycle Nth-stage feeding scheme. Using the normalized value of the deviation of the same attribute parameters of the first full-cycle first-stage feeding scheme and the second full-cycle first-stage feeding scheme as the dimensional deviation, the Euclidean distance is calculated to obtain the first-stage Euclidean distance. Until the normalized value of the deviation of the same attribute parameter between the first full-cycle Nth stage feeding scheme and the second full-cycle Nth stage feeding scheme is used as the dimensional deviation, the Euclidean distance is calculated to obtain the Nth stage Euclidean distance. Obtain the first stage weights up to the Nth stage weights based on the Delphi method, perform a weighted summation of the Euclidean distances from the first stage to the Nth stage, obtain the first feed feeding scheme deviation parameter, and add it to the set of feed feeding scheme deviation parameters.

5. The method as described in claim 1, characterized in that, Receive positioning information from the positioning beacon and configure the target area for the first bait placement, including: Obtain a 3D mesh model of a gravity-type wire cage; Extract the relative coordinate parameters of the bait receiving surface mesh model and the positioning beacon deployment position in the three-dimensional mesh model of the gravity cage. The bait receiving surface mesh model is the opening area on the water surface of the cage, which can receive bait. When receiving positioning information from the positioning beacon, the coordinate position of the bait receiving surface grid model is located using the positioning information as the origin and the relative coordinate parameters. Then, a world coordinate transformation is performed to obtain the first bait throwing target area.

6. The method as described in claim 1, characterized in that, Based on the drone's flight path, control the drone to execute the bait feeding plan, including: Collect wind speed and wind direction parameters in the target area where the bait is first scattered; Several bait throwing positions are evenly distributed at preset distances within a preset range above the first bait throwing target area; From the aforementioned feeding scheme, extract the feed density, feed particle size, and feeding amount; By using a feed waste prediction network, the network iterates through several feed placement locations and performs analysis based on the first feed placement target area, the wind speed parameter, the wind direction parameter, the feed density, the feed particle size, and the feeding amount to obtain several predicted feed waste values. Based on the feeding amount, the effective feeding amount is calculated by iterating through the predicted values ​​of several feed waste amounts; When the effective feeding amount of the aforementioned baits is greater than or equal to the effective feeding amount threshold, the corresponding bait scattering position is randomly selected, and the bait feeding plan is executed. When none of the predicted feed waste values ​​are greater than or equal to the effective feeding threshold, the feed scattering position is selected from the predicted feed waste values. The minimum value of the predicted feed amount and the predicted feed waste amount are summed to obtain an updated feed feeding plan, which is then executed at the minimum feed scattering location.

7. The method as described in claim 6, characterized in that, By using a feed waste prediction network, traversing several feed placement locations, and combining the first feed placement target area, wind speed parameters, wind direction parameters, feed density, feed particle size, and feeding amount, several predicted feed waste values ​​are obtained, including: Collect multiple sets of historical data on bait placement. Each set of these historical data includes data on bait placement location, bait placement target area, wind speed, wind direction, bait density, bait particle size, feeding amount, and bait waste. Using the recorded data on feed waste as supervision, and the recorded data on feed placement location, feed placement target area, wind speed, wind direction, feed density, feed particle size, and feed amount as input, a feed waste prediction network is trained using machine learning.

8. A drone-based precision baiting system for gravity-fed net cages, characterized in that, Applied to the fish breeding control terminal, for implementing the method as described in any one of claims 1 to 7, comprising: The positioning trigger module is used to receive positioning information from the positioning beacon and configure the first target area for baiting when the bait feeding clock associated with the positioning beacon number is triggered. The positioning beacon is deployed on the gravity net cage for raising adult fish. The path optimization module is used to obtain the drone's position and perform path optimization in conjunction with the target area of ​​the first bait placement to obtain the drone's flight path; The scheme matching module is used to input the positioning beacon number into the bait feeding scheme library and match the bait feeding scheme. The execution control module is used to control the drone to execute the bait feeding plan according to the drone's flight path.

Images