A method and system for controlling feeding of deep-water cage fish

Abstract

The present invention discloses a method and system for controlling feeding of deep-water cage fish, the method comprising generating a detailed list of farmed fish in a target deep-water area according to fish information, performing multiple rounds of pepper and salt denoising on the acquired image data of the target deep-water farm area, inputting underwater light data into a water body thermal stratification model to obtain each deep-water layer and short-wave radiation data corresponding to each deep-water layer; analyzing and identifying the image data through an AI image recognition algorithm to obtain fish growth data, constructing a feeding amount control model according to the detailed list of farmed fish, fish growth data and short-wave radiation data, and generating a farming visualization report to realize feeding control of fish in the target deep-water farm area. The present invention analyzes fish growth and health through image data, constructs and optimizes a feeding control model according to data such as ammonia nitrogen content, dissolved oxygen content and short-wave radiation, optimizes feeding amount control, so as to improve fish growth rate and quality, and realize long-term sustainable farming.

Description

Technical Field

[0001] The invention relates to the technical field of deep water aquaculture, and in particular to a deep water cage fish feeding control method and system. Background Art

[0002] With the continuous development of marine aquaculture technology, deep-water cage fish farming technology has emerged. This farming method uses large deep-water cages to farm fish in suitable sea areas. It has the advantages of strong wind and wave resistance, large farming capacity, fast fish growth rate, and good product quality.

[0003] The existing feeding method for deep-water cage aquaculture is manual timed and quantitative feeding. However, due to individual differences in fish, changes in growth rate and fluctuations in environmental factors, it is still difficult to achieve completely accurate feeding amounts. Excessive feeding can easily cause water quality deterioration, feed waste and affect fish health. Insufficient feeding can cause fish growth stunting, decreased immunity and even dead fish.

[0004] It can be seen that how to reasonably control the feeding of deep-water cage fish farming and promote the digitalization and intelligentization of deep-water cage farming has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0005] The invention provides a deep-water cage fish feeding control method and system, so as to carry out reasonable feeding control for deep-water cage culture.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a deep-water cage fish feeding control method, comprising:

[0007] The fish information of the target deep-water aquaculture area is obtained, and a detailed list of farmed fish in the target deep-water area is generated according to the fish information.

[0008] Multiple rounds of salt and pepper denoising are performed on the first image data of the target deep-water aquaculture area to obtain second image data, wherein each round of the salt and pepper denoising is designed to perform smoothing traversal and outlier removal operations on each of the first image data in turn through different window pixel frames, and each round of the salt and pepper denoising ignores the removal operation of local salt and pepper extreme points filtered by the window pixel frame.

[0009] The acquired underwater light data of the deep-water cages in the target deep-water aquaculture area is input into the water thermal stratification model constructed by DYRESM hydrodynamic technology to obtain the deep-water levels corresponding to the target deep-water aquaculture area and the short-wave radiation data corresponding to each of the deep-water levels.

[0010] The second image data is analyzed to obtain the fish growth data corresponding to each of the deep water layers, and a feeding amount control model is constructed according to the farmed fish list, the fish growth data and the shortwave radiation data.

[0011] An aquaculture visualization report is generated based on the output result of the feeding amount control model, and the feeding control of the fish in the target deep-water aquaculture area is achieved by using the aquaculture visualization report.

[0012] Furthermore, the window pixel frame at least includes a first window pixel frame and a second window pixel frame.

[0013] The step of performing multiple rounds of salt and pepper denoising processing on each first image data of the target deep-water aquaculture area to obtain second image data includes:

[0014] The first image data of the target deep-water aquaculture area is obtained. In the process of performing salt and pepper denoising on the first image data, a first smoothing traversal is performed on each of the first data images in turn according to the first window pixel frame to determine a candidate salt and pepper noise point set.

[0015] A second smoothing traversal is performed on each of the first data images in sequence according to the second window pixel frame, and the candidate salt and pepper noise point set is corrected based on the second smoothing traversal result, and local salt and pepper extreme points in the candidate salt and pepper noise are removed to obtain a target salt and pepper noise point set.

[0016] The target salt and pepper noise point set is eliminated from each of the first image data, and image reconstruction is performed based on the elimination result to obtain the second image data.

[0017] Furthermore, the feeding amount control model is constructed according to the farmed fish list, the fish growth data and the shortwave radiation data, including:

[0018] A feeding amount control model is constructed according to the farmed fish list and the fish growth data.

[0019] The water body parameters of the target deep-water aquaculture area are obtained, the ammonia nitrogen content and the dissolved oxygen content of the target deep-water aquaculture area are analyzed according to the water body parameters, a feeding amount adjustment coefficient is calculated according to the ammonia nitrogen content, the dissolved oxygen content and the shortwave radiation data, and the feeding amount control model is corrected according to the feeding amount adjustment coefficient.

[0020] The modified feeding amount control model is used to perform fish feeding simulation, and the feeding amount control model is iteratively optimized according to the feeding feedback until the feeding feedback reaches a preset standard.

[0021] Further, the feeding amount adjustment coefficient is expressed as:

[0022]

[0023] in, is the feeding amount adjustment coefficient, is the measured ammonia nitrogen content in the target deep-water aquaculture area, is the preset ammonia nitrogen threshold, is the tolerance range of ammonia nitrogen content, is the influence coefficient of ammonia nitrogen content on feeding amount, is the measured dissolved oxygen content in the target deep-water aquaculture area, is the preset dissolved oxygen threshold, is the tolerance range of dissolved oxygen, is the influence coefficient of dissolved oxygen on feeding amount, is the impact coefficient of shortwave radiation data in the target deep-water aquaculture area.

[0024] Further, the fish feeding simulation is performed using the modified feeding amount control model, and the feeding amount control model is iteratively optimized according to the feeding feedback until the feeding feedback reaches a preset standard, including:

[0025] A deep water simulation model of the target deep water aquaculture area is constructed according to the water body parameters.

[0026] The deepwater simulation model is subjected to continuous feeding simulation according to the feeding amount control model.

[0027] Obtain the difference in water quality change of the deepwater simulation model after each feeding simulation.

[0028] The deepwater simulation model is iteratively optimized according to the water quality change difference until the water quality change difference falls within a restricted change range.

[0029] Furthermore, the data analysis of the second image data to obtain the fish growth data corresponding to each of the deep water layers includes:

[0030] The first fish information in the dynamic image is acquired according to the second image data of the target deep-water aquaculture area.

[0031] Feature extraction is performed on underwater sonar images of the target deep-water aquaculture area at different time periods to obtain time-series static motion features of the target deep-water aquaculture area.

[0032] According to the time-series static motion characteristics, the first fish information is corrected to obtain the second fish information.

[0033] The fish growth status of the target deep-water aquaculture area is analyzed according to the second fish information to obtain fish growth data.

[0034] Furthermore, the obtaining of the first fish information in the dynamic image according to the second image data of the target deep-water aquaculture area includes:

[0035] An underwater dynamic image is obtained based on the second image data corresponding to adjacent acquisition times of the target deep-water aquaculture area.

[0036] The underwater dynamic image is recognized using an AI image recognition algorithm to obtain time-sharing fish feature data in the image.

[0037] An inter-frame feature difference analysis is performed on each adjacent time-sharing fish feature data to obtain first fish information.

[0038] Furthermore, the first fish information is corrected according to the time-series static motion characteristics to obtain the second fish information, including:

[0039] According to the temporal static motion characteristics, key frame position points and posture change points of the fish in the target deep-water breeding area are determined.

[0040] The first fish information is corrected in outline and repaired in deformation according to the key frame position points and the posture change points to obtain the second fish information.

[0041] Furthermore, the fish growth status in the target deep-water aquaculture area is analyzed according to the second fish information to obtain fish growth data, including:

[0042] The second fish information is matched with a fish growth information database to obtain growth parameter information of the corresponding fish.

[0043] A mapping relationship between the second fish information and the growth parameter information is constructed.

[0044] The growth status of fish in the target deep-water aquaculture area is evaluated according to the mapping relationship to obtain fish growth data.

[0045] Another embodiment of the present invention provides a deep-water cage fish feeding control system, comprising:

[0046] The information acquisition module is used to acquire fish information in the target deep-water aquaculture area and generate a detailed list of aquaculture fish in the target deep-water area based on the fish information.

[0047] The noise processing module is used to perform multiple rounds of salt and pepper denoising on the first image data of the target deep-water aquaculture area to obtain second image data, wherein each round of the salt and pepper denoising is designed to perform smoothing traversal and abnormal point removal operations on each of the first image data in turn through different window pixel frames, and each round of the salt and pepper denoising ignores the removal operation of local salt and pepper extreme points screened by the window pixel frame.

[0048] The radiation stratification module is used to input the acquired underwater light data of the deep-water cages in the target deep-water aquaculture area into the water body thermal stratification model constructed by DYRESM hydrodynamic technology, so as to obtain the deep-water levels corresponding to the target deep-water aquaculture area and the short-wave radiation data corresponding to each of the deep-water levels.

[0049] The model building module is used to perform data analysis on the second image data to obtain the corresponding fish growth data in each of the deep water layers, and to build a feeding amount control model based on the farmed fish list, the fish growth data and the shortwave radiation data.

[0050] A feeding control module is used to generate a farming visualization report based on the output results of the feeding amount control model, and to achieve feeding control of the fish in the target deep-water farming area with the farming visualization report.

[0051] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0052] By using high-speed computing and processing services, we can conduct intelligent analysis on the deep-water cage fish data obtained in various dimensions. By combining advanced image processing, model analysis and big data analysis technologies, we can realize the perception, judgment and analysis of deep-water cage fish and achieve ultimate feeding control, thereby optimizing the breeding environment, improving the utilization rate of bait, realizing refined management and intelligent decision-making of deep-water fish breeding, and promoting the development of fisheries in a more intelligent and sustainable direction.

[0053] By acquiring video images and sonar images of deep-water cage fish, analyzing the growth and health status of fish, and thus controlling the feeding of feed, the activity status and appearance of fish can be monitored in real time, and the feeding amount can be controlled based on the information obtained, thereby improving feed utilization and promoting healthy and rapid growth of fish. The feeding amount adjustment coefficient is calculated based on the ammonia nitrogen content and dissolved oxygen content in the water, so as to adjust the original feeding amount, reduce the residue and decomposition of feed in the water, reduce water pollution, improve the self-purification capacity of the water body, and achieve long-term sustainable breeding. Through multiple rounds of salt and pepper denoising, the image can be smoothly traversed and abnormal points can be eliminated through different window pixel frames, effectively removing salt and pepper noise and improving image quality. The water body thermal stratification model constructed using DYRESM hydrodynamic technology obtains each deep water level and the corresponding shortwave radiation data, so as to optimize and adjust the feeding amount control model according to the impact of water body thermal stratification on the fish growth environment, and improve the growth rate and quality of fish. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A flowchart of the steps of a deep-water cage fish feeding control method provided by an embodiment of the present invention;

[0055] Figure 2 A noise suppression effect diagram provided by an embodiment of the present invention;

[0056] Figure 3 A schematic diagram of water body stratification provided by an embodiment of the present invention;

[0057] Figure 4 This is a structural block diagram of a deep-water cage fish feeding control system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0060] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0061] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0062] An embodiment of the present invention provides a method for controlling feeding of deep-water cage fish. For details, see Figure 1 , Figure 1 The flowchart of the deep-water cage fish feeding control method in one embodiment of the present invention includes steps S11 to S15:

[0063] Step S11, obtaining fish information of the target deep-water aquaculture area, and generating a detailed list of aquaculture fish in the target deep-water area according to the fish information.

[0064] Deepwater aquaculture areas have great depths, complex environments, and a wide variety of fish species, and different fish species have different requirements for growth environments and feeds. Therefore, in order to scientifically manage aquaculture, it is first necessary to accurately obtain fish information in the target area.

[0065] Specifically, common deep-water farmed fish are shown in Table 1. The fish list includes at least information such as the scientific name of the fish, classification information, living habits, growth cycle and economic value.

[0066] Table 1 Deep water cultured fish

[0067]

[0068] Generating a detailed list of farmed fish can help farmers clearly understand the key information such as the types, quantity, growth cycle, etc. of fish in the farming area, and provide data support for subsequent management and feeding control. At the same time, the detailed list can also be used as part of the farming record, which is convenient for farmers to track and analyze the farming results.

[0069] Step S12, performing multiple rounds of salt and pepper denoising processing on each first image data of the acquired target deep-water aquaculture area to obtain second image data, wherein each round of salt and pepper denoising processing is designed to perform smoothing traversal and abnormal point removal operations on each first image data in turn through different window pixel frames, and each round of salt and pepper denoising processing ignores the removal operation of local salt and pepper extreme points screened by the window pixel frame.

[0070] For deep-water images of deep-water aquaculture areas, due to poor lighting conditions, deep-water images often show small color differences and single colors (only black and white blocks), which leads to the presence of more black (pepper) and white (salt) pixels in deep-water images, which significantly reduces the quality and visualization of deep-water images. In addition, deep-water images also need to be acquired by an image acquisition device, and the image acquisition device, whether it uses wired communication or wireless communication, will be affected by long-distance transmission in deep water, which further aggravates the salt and pepper noise phenomenon. In view of this, the embodiment of the present invention adopts the following related technical means to achieve the removal of salt and pepper noise points in deep-water images, so that the details and features in the image are easy to identify, ensuring the accuracy and reliability of the image, and providing strong support for subsequent scientific research, detection, aquaculture and other activities.

[0071] The image data acquired underwater is mainly affected by salt and pepper noise and Gaussian noise due to the influence of equipment and environment. Therefore, it is necessary to perform salt and pepper denoising on the first image data of the target deep-water aquaculture area. Specifically, since the median filter can filter out random pulse noise in the image while retaining the edge information of the image, this embodiment uses the median filtering method to remove salt and pepper noise.

[0072] like Figure 2 As shown, Figure 2 The figure shows the noise suppression effect of the deep-water cage fish feeding control method provided by the embodiment of the present invention, Figure 2 (a) is the first image data diagram, Figure 2 (b) is a grayscale image of the first image data, Figure 2 (c) Image after single salt and pepper filtering. Figure 2 (d) is the image after multiple rounds of salt and pepper filtering. Figure 2(e) is the image after Gaussian filtering. Specifically, when filtering the first image data, it is necessary to select a suitable first window pixel frame according to the image pixels. This embodiment takes the filtering window size of 3×3 as an example, sequentially traverses each acquired first image data and calculates the grayscale value of the center pixel in the window. When the grayscale value of the center pixel is equal to the maximum grayscale value or the minimum grayscale value in the first window pixel frame, the center point of the window may be a noise point, and the point is marked as a candidate salt and pepper noise point.

[0073] However, the candidate salt and pepper noise point is a local extreme point in a 3×3 window size. If there are multiple local extreme points in the window, it may cause misjudgment of the noise point. Therefore, in each round of salt and pepper denoising in this embodiment, after obtaining the candidate salt and pepper noise point, it is necessary to adjust the size of the window pixel frame and use the second window pixel frame to further detect and confirm the candidate salt and pepper noise point in the image.

[0074] Specifically, the size of the second window pixel frame should be slightly larger than the first window pixel frame, such as 5×5, 7×7, 9×9 and other window sizes. The second window pixel frame is used to traverse the first window pixel frame again. If the grayscale value of the central pixel of the candidate salt and pepper noise point in the second window pixel frame is still equal to the maximum grayscale value or the minimum grayscale value in the second window pixel frame, the point is confirmed to be a salt and pepper noise point and is removed.

[0075] Due to the complexity of the information in the underwater image, it is difficult to achieve a good filtering effect with a single filtering denoising. Therefore, this embodiment records the removed noise points each time the filtering is performed. If the number of noise points removed by the current filtering is not 0, the output of the median filter is used as the input of the next filtering operation, and the above filtering operation is performed again until the number of noise points removed by a single filtering drops to 0 or the number of filtering times reaches a preset maximum number. At this time, the output of the median filter is obtained as the second image data. By using window pixel frames of different sizes for multiple rounds of salt and pepper denoising, the accuracy of noise detection can be significantly improved, image details can be protected, image quality can be enhanced, and the robustness of the algorithm can be improved.

[0076] Preferably, after multiple rounds of salt and pepper denoising, in order to further improve the image quality and remove Gaussian noise in the image, the image after multiple rounds of salt and pepper denoising can also be subjected to Gaussian filtering through a Gaussian filter. The specific filtering effect is as follows: Figure 2 (e) as shown.

[0077] Multiple rounds of salt and pepper denoising can smoothly traverse the image and remove abnormal points through different window pixel frames, effectively removing salt and pepper noise and improving image quality. At the same time, each round of processing ignores the removal of salt and pepper extreme points, which can avoid accidentally deleting important image information. The processed image data is more accurate and clear, which is conducive to subsequent analysis and judgment.

[0078] Step S13, inputting the acquired underwater light data of the deep-water cages in the target deep-water aquaculture area into a water thermal stratification model constructed by the DYRESM (Dynamic Reservoir Simulation Model) hydrodynamic technology, to obtain the various deep-water layers corresponding to the target deep-water aquaculture area and the short-wave radiation data corresponding to each deep-water layer.

[0079] Underwater light data in deep-water aquaculture areas has an important impact on fish growth and feeding control. At the same time, thermal stratification of water bodies will also affect environmental factors such as water temperature and dissolved oxygen, which in turn affects fish growth. Figure 3 As shown, different deep-water farmed fish can adapt to a certain range of deep-water pressure. Therefore, the short-wave radiation brought by light varies with the increase of water depth. Different light intensities will have different effects on fish growth. Therefore, it is necessary to obtain underwater light data and consider the influence of water thermal stratification.

[0080] Use professional underwater photometers or underwater spectrometers to measure at different depths in the target deep-water aquaculture area and obtain underwater light data at each deep-water level. These data should include parameters such as the intensity and distribution of light of different wavelengths.

[0081] A water thermal stratification model suitable for the target deep-water aquaculture area was constructed using DYRESM hydrodynamic technology. The water thermal stratification model was used to simulate the vertical distribution of parameters such as temperature, salinity, and density in the water body and their changes over time.

[0082] The acquired underwater light data will be input into the constructed water body thermal stratification model. The underwater light data will be used as part of the model input to simulate the propagation, absorption and scattering of light in the water body, thereby affecting the thermal stratification structure of the water body.

[0083] Run the water thermal stratification model to simulate the water thermal stratification structure of the target deep-water aquaculture area. During the simulation, the model will consider the impact of underwater light data on water thermal stratification and calculate the shortwave radiation data corresponding to each deep-water level.

[0084] By acquiring underwater light data and inputting it into the water thermal stratification model, the water thermal stratification structure of the target deep-water aquaculture area can be simulated more accurately, which helps to gain a deeper understanding of the thermal stratification characteristics of the water body and its impact on deep-water aquaculture organisms. At the same time, the simulated short-wave radiation data corresponding to each deep-water layer can be used to evaluate the impact of underwater light conditions on deep-water aquaculture organisms. Based on these data, aquaculture strategies can be adjusted, such as selecting appropriate aquaculture depths, optimizing light conditions, etc., to improve the biomass and quality of deep-water aquaculture.

[0085] In addition, by gaining a deep understanding of the thermal stratification characteristics of water bodies and the impact of underwater lighting conditions on deep-water aquaculture organisms, we can provide a scientific basis and technical support for the development of the deep-water aquaculture industry, which will help promote the sustainable development and industrial upgrading of the deep-water aquaculture industry.

[0086] Step S14, performing data analysis on the second image data to obtain the corresponding fish growth data in each deep water layer, and constructing a feeding amount control model based on the farmed fish list, fish growth data and shortwave radiation data.

[0087] Reasonable feeding amount is the key to ensure the healthy growth of fish and improve the breeding efficiency. By building a model, the feeding amount can be accurately controlled to avoid overfeeding or underfeeding. Therefore, it is necessary to build an initial feeding amount control model based on the farmed fish list and fish growth data.

[0088] The second image data corresponding to the first image data acquired at adjacent acquisition time points constitute a continuous underwater dynamic image sequence. According to the second image data corresponding to the adjacent acquisition time of the target deep-water aquaculture area, an underwater dynamic image is obtained. The underwater dynamic image can intuitively reflect the activity status of fish in the aquaculture area and is an important source of fish information. By continuously acquiring image data, the dynamic change process of fish can be captured.

[0089] AI image recognition algorithms are used to identify underwater dynamic images to obtain the time-sharing fish feature data in the images. The feature data includes the type, quantity, size, swimming speed, direction, etc. of the fish.

[0090] The inter-frame feature difference analysis is performed on each adjacent time-sharing fish feature data to compare the changes in fish features at different time points. Through the difference analysis, the dynamic behavior information of the fish, such as swimming trajectory, speed change, etc., can be obtained to obtain the first fish information.

[0091] Feature extraction is performed on underwater sonar images of the target deep-water aquaculture area at different times to obtain the time-series static motion characteristics of the target deep-water aquaculture area. Underwater sonar images can penetrate the water body and capture the static position and posture information of fish, which is crucial for understanding the behavior patterns and growth status of fish.

[0092] According to the time-series static motion characteristics, the key frame position points and posture change points of fish at different time periods are determined. These points represent the position and posture information of fish at specific time points, which are important bases for analyzing the behavior patterns and growth status of fish. They can accurately reflect the changes in the behavior patterns and growth status of fish, and help to more accurately evaluate the health status and growth potential of fish.

[0093] According to the key frame position points and posture change points, the first fish information is corrected for contour correction and deformation repair. Through comparison and analysis, the errors or deformations that may occur during image acquisition and processing are corrected to obtain more accurate and complete second fish information, reducing the impact of errors or deformations on fish information during image acquisition and processing.

[0094] The second fish information is matched with the fish growth information database to obtain the growth parameter information of the corresponding fish, and the growth parameter information includes the growth rate, feed utilization rate, health status, etc. of the fish. The mapping relationship between the second fish information and the growth parameter information is constructed to reflect the internal connection and law between the fish information and the growth parameters, and provide a scientific basis for the subsequent growth status evaluation, which is used to evaluate the growth status and health status of the fish.

[0095] Water parameters directly affect the growth environment and health of fish. Excessive ammonia nitrogen content and low dissolved oxygen content will have adverse effects on fish, so these parameters need to be monitored and analyzed in real time. Water quality monitoring equipment is used to obtain water parameters of the target deep-water aquaculture area, including ammonia nitrogen content, dissolved oxygen content, etc. Ammonia nitrogen content and dissolved oxygen content are analyzed based on these parameters to assess the health of the water body.

[0096] Changes in water parameters and shortwave radiation data will affect the appetite and digestion capacity of fish, thus affecting the feeding amount requirements. By calculating the adjustment coefficient, the feeding amount can be dynamically adjusted to adapt to changes in environmental parameters, obtain the water parameters of the target deep-water aquaculture area, analyze the ammonia nitrogen content and dissolved oxygen content of the target deep-water aquaculture area according to the water parameters, calculate the feeding amount adjustment coefficient according to the ammonia nitrogen content, dissolved oxygen content and shortwave radiation data, and modify the feeding amount control model according to the feeding amount adjustment coefficient.

[0097] The revised feeding amount control model is used to simulate fish feeding, and the feeding amount control model is iteratively optimized according to the feeding feedback until the feeding feedback reaches the preset standard.

[0098] The feed amount adjustment coefficient is expressed as:

[0099]

[0100] in, is the feeding amount adjustment coefficient, is the measured ammonia nitrogen content in the target deep-water aquaculture area, is the preset ammonia nitrogen threshold, is the tolerance range of ammonia nitrogen content, is the influence coefficient of ammonia nitrogen content on feeding amount, is the measured dissolved oxygen content in the target deep-water aquaculture area, is the preset dissolved oxygen threshold, is the tolerance range of dissolved oxygen, is the influence coefficient of dissolved oxygen on feeding amount, is the shortwave radiation data impact coefficient of the target deep-water aquaculture area. By using the feeding amount adjustment coefficient to correct the feeding amount control model, it can more accurately reflect the feeding amount requirements in the actual aquaculture environment and improve the practicality and accuracy of the model.

[0101] Specifically, the present embodiment uses the modified feeding amount control model to perform fish feeding simulation, and iteratively optimizes the feeding amount control model according to the feeding feedback until the feeding feedback reaches the preset standard. The process is:

[0102] A deep-water simulation model of the target deep-water aquaculture area is constructed based on water body parameters. The water quality and environmental conditions in deep-water aquaculture areas are complex and changeable. Constructing a simulation model helps to accurately simulate and predict the changing trends of various factors in aquaculture water bodies, and provide basic data support for the optimization of feeding amount control models.

[0103] Changes in feeding amount will directly affect water quality indicators in aquaculture water, such as ammonia nitrogen content, dissolved oxygen content, etc. According to the feeding amount control model, continuous feeding simulation is performed on the deep water simulation model. Through feeding simulation, the impact of different feeding amounts on water quality can be evaluated, providing a basis for the optimization of the feeding amount control model.

[0104] After each feeding simulation, the water quality indicators in the simulation model are compared with the preset water quality standards or historical data, and the difference in water quality changes is calculated. The difference reflects the impact of the feeding simulation on the water quality. The difference in water quality changes in the deepwater simulation model after each feeding simulation is obtained. The difference in water quality changes is an important indicator for evaluating the effect of the feeding amount control model. By calculating the difference, the impact of the feeding amount on water quality can be quantified, providing data support for the iterative optimization of the model.

[0105] By continuously adjusting and optimizing the model parameters, the simulation model can more accurately reflect the water quality change process in the actual aquaculture environment. The deep-water simulation model is iteratively optimized according to the water quality change difference. During the optimization process, the relevant parameters and boundary conditions in the model are adjusted to reduce the water quality change difference to within the restricted range.

[0106] Step S15, generating a farming visualization report based on the output result of the feeding amount control model, and realizing feeding control of fish in the target deep-water farming area by using the farming visualization report.

[0107] The feeding amount control model outputs a reasonable feeding amount based on the type, quantity, size, growth status, health status, etc. of fish information in the target deep-water aquaculture area, as well as environmental factors such as water temperature, water quality, and light.

[0108] The output results of the feeding amount control model are converted into an easy-to-understand and easy-to-operate farming visualization report, which intuitively displays information such as feeding recommendations, fish growth trends, and changes in environmental factors.

[0109] The aquaculture visualization report is presented in a combination of charts and text. It can use large language model technology to generate natural language to provide reference for aquaculture personnel. It can also dynamically adjust the content of the report according to the user's newly added fish and record the content of the report after each adjustment to facilitate vertical comparison. This allows you to understand the changes in feeding adjustments after adding new fish, and then understand the ecological environment in the deep-water fish tank.

[0110] According to the farming visualization report, farmers can intuitively understand the feeding recommendations and fish growth status, so as to implement feeding control, adjust the settings of the feeder, select the appropriate feed type, monitor the feeding process, etc. Farmers can also flexibly adjust the feeding strategy according to the actual situation and the status of the farmed fish presented in the report to cope with various changes in the growth process of the fish.

[0111] The deep-water cage fish feeding control method of the present invention acquires video images and sonar images of deep-water cage fish, analyzes the growth status and health status of the fish, thereby controlling the feeding of feed, can monitor the activity status and appearance of the fish in real time, and control the feeding amount according to the obtained information, thereby improving the feed utilization rate and promoting the healthy and rapid growth of the fish; the feeding amount adjustment coefficient is calculated according to the ammonia nitrogen content and the dissolved oxygen content in the water, thereby adjusting the original feeding amount, reducing the residue and decomposition of the feed in the water, reducing the pollution to the water quality, improving the self-purification capacity of the water body, and realizing long-term sustainable breeding; through multiple rounds of pepper and salt denoising processing, the image can be smoothly traversed and abnormal points can be eliminated through different window pixel frames, so as to effectively remove the pepper and salt noise and improve the image quality; the water body thermal stratification model constructed by using the DYRESM hydrodynamic technology obtains each deep water level and the corresponding shortwave radiation data, so as to optimize and adjust the feeding amount control model according to the influence of the water body thermal stratification on the fish growth environment, and improve the growth rate and quality of the fish.

[0112] The embodiment of the present invention further provides a deep-water cage fish feeding control system, which is used to execute the deep-water cage fish feeding control method as described above. Figure 4This is a structural block diagram of a deep-water cage fish feeding control system according to an embodiment of the present invention. The system comprises:

[0113] The information acquisition module 21 is used to acquire the fish information in the target deep-water aquaculture area, and generate a detailed list of aquaculture fish in the target deep-water area according to the fish information.

[0114] The noise processing module 22 is used to perform multiple rounds of salt and pepper denoising on the first image data of the target deep-water aquaculture area to obtain second image data, wherein each round of the salt and pepper denoising is designed to perform smoothing traversal and abnormal point removal operations on each of the first image data in turn through different window pixel frames, and each round of the salt and pepper denoising ignores the removal operation of local salt and pepper extreme points filtered by the window pixel frame.

[0115] The radiation stratification module 23 is used to input the acquired underwater light data of the deep-water cages in the target deep-water aquaculture area into the water body thermal stratification model constructed by DYRESM hydrodynamic technology, so as to obtain the various deep-water layers corresponding to the target deep-water aquaculture area and the short-wave radiation data corresponding to each of the deep-water layers.

[0116] The model building module 24 is used to perform data analysis on the second image data to obtain the corresponding fish growth data in each of the deep water layers, and to build a feeding amount control model based on the farmed fish list, the fish growth data and the shortwave radiation data.

[0117] The feeding control module 25 is used to generate a farming visualization report based on the output result of the feeding amount control model, and to achieve feeding control of the fish in the target deep-water farming area with the farming visualization report.

[0118] The technical features and technical effects of the system proposed in the embodiment of the present invention are the same as those of the method proposed in the embodiment of the present invention, and are not described in detail here. Each module in the above system can be implemented in whole or in part by software, hardware, and a combination thereof. Each of the above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each of the above modules.

[0119] In summary, the deep-water cage fish feeding control method and system provided by the embodiment of the present invention has the following beneficial effects compared with the prior art:

[0120] By using high-speed computing and processing services, we can conduct intelligent analysis on the deep-water cage fish data obtained in various dimensions. By combining advanced image processing, model analysis and big data analysis technologies, we can realize the perception, judgment and analysis of deep-water cage fish and achieve ultimate feeding control, thereby optimizing the breeding environment, improving the utilization rate of bait, realizing refined management and intelligent decision-making of deep-water fish breeding, and promoting the development of fisheries in a more intelligent and sustainable direction.

[0121] By acquiring video images and sonar images of deep-water cage fish, analyzing the growth and health status of fish, and thus controlling the feeding of feed, the activity status and appearance of fish can be monitored in real time, and the feeding amount can be controlled based on the information obtained, thereby improving feed utilization and promoting healthy and rapid growth of fish. The feeding amount adjustment coefficient is calculated based on the ammonia nitrogen content and dissolved oxygen content in the water, so as to adjust the original feeding amount, reduce the residue and decomposition of feed in the water, reduce water pollution, improve the self-purification capacity of the water body, and achieve long-term sustainable breeding. Through multiple rounds of salt and pepper denoising, the image can be smoothly traversed and abnormal points can be eliminated through different window pixel frames, effectively removing salt and pepper noise and improving image quality. The water body thermal stratification model constructed using DYRESM hydrodynamic technology obtains each deep water level and the corresponding shortwave radiation data, so as to optimize and adjust the feeding amount control model according to the impact of water body thermal stratification on the fish growth environment, and improve the growth rate and quality of fish.

[0122] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for controlling feeding of deep-water cage fish, characterized in that: include: Acquire fish information of a target deep-water aquaculture area, and generate a detailed list of farmed fish in the target deep-water aquaculture area according to the fish information; Perform multiple rounds of salt and pepper denoising on each first image data of the target deep-water aquaculture area to obtain second image data, wherein each round of the salt and pepper denoising is designed to sequentially perform smoothing traversal and outlier removal operations on each first image data through different window pixel frames, and each round of the salt and pepper denoising ignores the removal operation of local salt and pepper extreme value points screened by the window pixel frame; Input the acquired underwater light data of the deep-water cages in the target deep-water aquaculture area into the water body thermal stratification model constructed by DYRESM hydrodynamic technology to obtain the various deep-water layers corresponding to the target deep-water aquaculture area and the short-wave radiation data corresponding to each of the deep-water layers; The second image data is analyzed to obtain the corresponding fish growth data in each deep water layer, and a feeding amount control model is constructed according to the farmed fish list, the fish growth data and the shortwave radiation data; specifically comprising: constructing a feeding amount control model according to the farmed fish list and the fish growth data; obtaining water body parameters of the target deep water aquaculture area, analyzing the ammonia nitrogen content and dissolved oxygen content of the target deep water aquaculture area according to the water body parameters, calculating a feeding amount adjustment coefficient according to the ammonia nitrogen content, the dissolved oxygen content and the shortwave radiation data, and correcting the feeding amount control model according to the feeding amount adjustment coefficient; using the corrected feeding amount control model to simulate fish feeding, and iteratively optimizing the feeding amount control model according to feeding feedback until the feeding feedback reaches a preset standard; the feeding amount adjustment coefficient is expressed as: ； in, is the feeding amount adjustment coefficient, is the measured ammonia nitrogen content in the target deep-water aquaculture area, is the preset ammonia nitrogen threshold, is the tolerance range of ammonia nitrogen content, is the influence coefficient of ammonia nitrogen content on feeding amount, is the measured dissolved oxygen content in the target deep-water aquaculture area, is the preset dissolved oxygen threshold, is the tolerance range of dissolved oxygen, is the influence coefficient of dissolved oxygen on feeding amount, is the shortwave radiation data impact coefficient of the target deep-water aquaculture area; An aquaculture visualization report is generated based on the output result of the feeding amount control model, and the feeding control of the fish in the target deep-water aquaculture area is achieved by using the aquaculture visualization report.

2. The deep-water cage fish feeding control method according to claim 1, characterized in that: The window pixel frame at least includes a first window pixel frame and a second window pixel frame; The step of performing multiple rounds of salt and pepper denoising processing on each first image data of the target deep-water aquaculture area to obtain second image data includes: Acquire first image data of the target deep-water aquaculture area, and in the process of performing salt and pepper denoising on the first image data, perform a first smoothing traversal on each of the first image data in turn according to the first window pixel frame to determine a candidate salt and pepper noise point set; performing a second smoothing traversal on each of the first image data in sequence according to the second window pixel frame, modifying the candidate salt and pepper noise point set based on the second smoothing traversal result, removing local salt and pepper extreme value points in the candidate salt and pepper noise, and obtaining a target salt and pepper noise point set; The target salt and pepper noise point set is eliminated from each of the first image data, and image reconstruction is performed based on the elimination result to obtain the second image data.

3. The deep-water cage fish feeding control method according to claim 1, characterized in that: The method of using the modified feeding amount control model to simulate fish feeding, and iteratively optimizing the feeding amount control model according to feeding feedback until the feeding feedback reaches a preset standard, includes: Constructing a deep water simulation model of the target deep water aquaculture area according to the water body parameters; Performing continuous feeding simulation on the deepwater simulation model according to the feeding amount control model; Obtaining the difference in water quality change of the deepwater simulation model after each feeding simulation; The deepwater simulation model is iteratively optimized according to the water quality change difference until the water quality change difference falls within a restricted change range.

4. The deep-water cage fish feeding control method according to claim 1, characterized in that: The performing of data analysis on the second image data to obtain corresponding fish growth data in each of the deep water layers includes: acquiring first fish information in the second image data according to the second image data of the target deep-water aquaculture area; Extracting features from underwater sonar images of the target deep-water aquaculture area at different time periods to obtain time-series static motion features of the target deep-water aquaculture area; According to the time-series static motion characteristics, the first fish information is corrected to obtain second fish information; The fish growth status of the target deep-water aquaculture area is analyzed according to the second fish information to obtain fish growth data.

5. The deep-water cage fish feeding control method according to claim 4, characterized in that: The step of acquiring first fish information in the second image data according to the second image data of the target deep-water aquaculture area includes: Obtain an underwater dynamic image according to the second image data corresponding to adjacent acquisition times of the target deep-water aquaculture area; The underwater dynamic image is recognized by using an AI image recognition algorithm to obtain time-sharing fish feature data in the image; An inter-frame feature difference analysis is performed on each adjacent time-sharing fish feature data to obtain first fish information.

6. The deep-water cage fish feeding control method according to claim 4, characterized in that: The step of correcting the first fish information according to the time-series static motion characteristics to obtain the second fish information includes: Determining key frame position points and posture change points of fish in the target deep-water aquaculture area according to the time-series static motion characteristics; The first fish information is corrected in outline and repaired in deformation according to the key frame position points and the posture change points to obtain the second fish information.

7. The deep-water cage fish feeding control method according to claim 4, characterized in that: The step of analyzing the fish growth status of the target deep-water aquaculture area according to the second fish information to obtain fish growth data includes: Matching the second fish information with a fish growth information database to obtain growth parameter information of the corresponding fish; Constructing a mapping relationship between the second fish information and the growth parameter information; The growth status of fish in the target deep-water aquaculture area is evaluated according to the mapping relationship to obtain fish growth data.

8. A deep-water cage fish feeding control system, characterized in that: include: An information acquisition module is used to acquire fish information in a target deep-water aquaculture area, and generate a detailed list of farmed fish in the target deep-water aquaculture area according to the fish information; A noise processing module, used for performing multiple rounds of salt and pepper denoising on each first image data of the target deep-water aquaculture area to obtain second image data, wherein each round of the salt and pepper denoising is designed to sequentially perform smoothing traversal and outlier removal operations on each first image data through different window pixel frames, and each round of the salt and pepper denoising ignores the removal operation of local salt and pepper extreme value points screened by the window pixel frame; A radiation stratification module is used to input the acquired underwater light data of the deep-water cages in the target deep-water aquaculture area into a water thermal stratification model constructed by DYRESM hydrodynamic technology to obtain each deep-water layer corresponding to the target deep-water aquaculture area and short-wave radiation data corresponding to each deep-water layer; A model building module is used to perform data analysis on the second image data, obtain the corresponding fish growth data in each of the deep water layers, and build a feeding amount control model according to the farmed fish list, the fish growth data and the shortwave radiation data; specifically comprising: building a feeding amount control model according to the farmed fish list and the fish growth data; obtaining water body parameters of the target deep water aquaculture area, analyzing the ammonia nitrogen content and dissolved oxygen content of the target deep water aquaculture area according to the water body parameters, calculating a feeding amount adjustment coefficient according to the ammonia nitrogen content, the dissolved oxygen content and the shortwave radiation data, and correcting the feeding amount control model according to the feeding amount adjustment coefficient; using the corrected feeding amount control model to simulate fish feeding, and iteratively optimizing the feeding amount control model according to feeding feedback until the feeding feedback reaches a preset standard; the feeding amount adjustment coefficient is expressed as: ； in, is the feeding amount adjustment coefficient, is the measured ammonia nitrogen content in the target deep-water aquaculture area, is the preset ammonia nitrogen threshold, is the tolerance range of ammonia nitrogen content, is the influence coefficient of ammonia nitrogen content on feeding amount, is the measured dissolved oxygen content in the target deep-water aquaculture area, is the preset dissolved oxygen threshold, is the tolerance range of dissolved oxygen, is the influence coefficient of dissolved oxygen on feeding amount, is the shortwave radiation data impact coefficient of the target deep-water aquaculture area; A feeding control module is used to generate a farming visualization report based on the output results of the feeding amount control model, and to achieve feeding control of the fish in the target deep-water farming area with the farming visualization report.

Images