Intelligent training bait and health management method suitable for land-sea relay culture of cobia

Abstract

This invention relates to the field of cobia farming, and discloses an intelligent feeding and health management method suitable for cobia relay farming across land and sea. The method includes the following steps: healthy cobia relay farming of fry is achieved through feeding screening and classification, transportation stress assessment, long-distance transportation technology optimization, water quality parameter control, and intelligent feeding. A superior farming environment is provided by using wind- and wave-resistant marine cages. This invention ensures the nutritional needs of the fry. Simultaneously, scientific farming management and timely disease observation and prevention guarantee the healthy and rapid growth of the fry.

Description

Technical Field

[0001] This invention relates to the field of cobia farming, and in particular to intelligent feeding and health management methods applicable to cobia farming in both land and sea. Background Technology

[0002] Cobia, also known as sea bream or sea dragonfish, belongs to the order Perciformes, family Cobiaidae, and genus Cobia. It grows extremely fast, is rich in nutrients, and has strong disease resistance, making it a highly valuable species in marine aquaculture. In recent years, significant breakthroughs have been made in the artificial breeding and offshore cage culture techniques of cobia, opening up new avenues for increasing the income and wealth of coastal fishermen.

[0003] Because the cobia fry received are averaging 4-5cm in size and are relatively weak, they need to be transported long distances after being harvested from ponds, eventually reaching sea-resistant net cages for growth and rearing. This cross-regional and cross-environmental transport poses a severe test to the physiological tolerance of the fry. During the initial transportation phase, fry losses are significant, with an initial survival rate of only about 70%. At this stage, the fry are still feeding on live food and have not yet been acclimated to artificial feed. The acclimation process after arriving at the sea cages is also a period of high fry loss, with initial acclimation losses accounting for about 20% of the total fry. This indicates that transportation and acclimation are the two core challenges of this project and the key factors leading to the high initial mortality rate. Therefore, it is necessary to propose an intelligent feeding and health management method suitable for cobia's land-sea relay aquaculture, focusing on and optimizing core aspects such as fry health assessment and selection, long-distance transportation technology optimization, and feeding management during the sea-based growth and rearing period. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an intelligent feeding and health management method suitable for the land-sea relay farming of cobia.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides an intelligent feeding and health management method for cobia in land-sea relay aquaculture, comprising the following steps: Feeding behavior analysis was conducted on cobia fry to generate a feeding activity score, and the cobia fry were screened and classified based on the score results; A transportation stress assessment was conducted on each batch of target cobia fry, and the stress resistance of the target cobia fry was strengthened based on the assessment results, while water quality parameters were controlled during transportation. Adaptive feeding training was conducted on cobia fry in the target transport cage, and image analysis was combined during the feeding training process to formulate multimodal feeding strategies. During the transportation of the target cobia fry through the target transport cages, intestinal health early warning and functional feed intervention were carried out simultaneously for each batch of target cobia fry.

[0006] Furthermore, in a preferred embodiment of the present invention, the step of analyzing the feeding behavior of cobia fry to generate a feeding activity score for the cobia fry, and then screening and classifying the cobia fry based on the score results, specifically involves: Identify the current holding pond for cobia fry and designate it as the target holding pond, then install cameras in the target holding pond; Different batches of cobia fry were starved in the target holding tank. After starvation, live food was put into the target holding tank. At the same time, all the cameras were turned on to record real-time videos of the same batch of cobia fry feeding in the target holding tank, which were labeled as feeding behavior analysis videos. The feeding behavior analysis video is preprocessed, including video noise reduction and video interpolation optimization, to obtain a preprocessed feeding behavior analysis video. A trajectory tracking algorithm was introduced to perform trajectory tracking analysis on the same batch of cobia fry in pre-processed feeding behavior analysis videos, and to record the feeding behavior characteristics of the cobia fry. The feeding behavior characteristics of the cobia fry include reaction agility characteristics, feeding enthusiasm characteristics, competitive behavior characteristics, and feeding focus characteristics. A big data network is introduced to retrieve a feeding behavior characteristic scoring map. The feeding behavior characteristic scoring map records the scoring values ​​of different feeding behavior characteristics and presets a feeding enthusiasm scoring threshold. Among them, the feeding enthusiasm scoring threshold divides the fry of Cobra scad into an elimination group and a qualified group. The feeding behavior characteristics of cobia fry are imported into the feeding behavior characteristic scoring map for self-scoring. The feeding enthusiasm score of different batches of cobia fry is output. Combined with the feeding enthusiasm score threshold, the batches of cobia fry in the qualified group are marked as the target cobia fry batches.

[0007] Furthermore, in a preferred embodiment of the present invention, the step of conducting a transport stress assessment on batches of target cobia fry and strengthening the stress resistance of the target cobia fry based on the assessment results, while controlling water quality parameters during transport, specifically includes: In the target holding tank, batches of target cobia fry were sampled and their blood oxygen saturation and behavioral stress were monitored. The behavioral stress monitoring tail subjected the sampled target cobia fry to stress treatment and recorded the changes in swimming speed of the samples after the stress treatment. The study analyzes the changes in swimming speed of samples after stress treatment, extracts the time when the swimming speed of samples drops to a preset threshold after stress treatment, calibrates it as the sample recovery time, and outputs the stress analysis status of the target cobia batch by combining the blood oxygen saturation of the samples. The planned transportation time is predetermined, and a transportation emergency vulnerability algorithm model is introduced. The transportation emergency vulnerability algorithm model combines the stress analysis status of the target cobia batch, the feeding enthusiasm score of the target cobia batch, and the planned transportation time to calculate the transportation emergency vulnerability index of the target cobia batch. By retrieving the nutritional fortification knowledge graph through big data networks, the transportation emergency vulnerability index of the target cobia batch is mapped to the nutritional fortification knowledge graph, and different levels of nutritional supplementation schemes corresponding to different transportation emergency vulnerability indices are marked in the nutritional fortification knowledge graph. In the target holding tank, based on the transportation emergency vulnerability index of the target cobia batch, a corresponding nutritional supplementation plan is output. In the process of outputting the corresponding nutritional supplementation plan, the feeding equipment is used to feed the fish in a quantitative manner that affects the supplementation plan. The system acquires the transport cages for the target batch of cobia and simultaneously monitors and controls the water quality parameters of the transport cages during the transport process.

[0008] Furthermore, in a preferred embodiment of the present invention, the step of acquiring the transport cages for the target batch of cobia and simultaneously monitoring and controlling the water quality parameters of the transport cages during the transport of the target cobia batch specifically includes: The transport cages for the target batch of cobia are identified and designated as target transport cages. After outputting the corresponding nutrient addition plan, the target batch of cobia is transferred from the target holding pond to the target transport cage. The target transport cage is equipped with a sensor network for real-time monitoring of water quality parameters. An adaptive feeding system is used to perform preliminary feeding on batches of target cobia fry transferred from the target holding tank to the target transport cage during the transfer process. The preliminary feeding process involves intelligently feeding the target transport cage according to the adaptive feeding system and controlling the amount of feed fed to be equal to the amount fed in the target holding tank. In the target transport cage, the water quality parameters of the target cobia batch are recorded immediately after transfer. The transfer analysis time is preset, and the changes in the water quality parameters of the target cobia batch during the transfer analysis time are recorded. Combined with calculations, the trend of water quality parameter changes in the target transport cage is generated. Intelligent water quality parameter adjustment equipment is introduced into the target transport cage. The trend of water quality parameter changes in the target transport cage is imported into the intelligent water quality parameter adjustment equipment, and a closed-loop intervention mechanism is set in the intelligent water quality parameter adjustment equipment. The closed-loop intervention mechanism involves setting an adjustment threshold for the water quality parameters that need to be intervened and adjusted based on the changing trend of the water quality parameters of the target transport cage, and controlling the intelligent water quality parameter adjustment device to control the water quality parameters when the changing trend of the water quality parameters of the target transport cage does not remain at the adjustment threshold, so that the water quality parameters are always maintained within the standard range.

[0009] Furthermore, in a preferred embodiment of the present invention, the adaptive feeding training of cobia fry in the target transport cage, and the combination of image analysis during the feeding training process to formulate a multimodal feeding strategy, specifically includes: In the target transport cage, underwater cameras and feeding devices are installed, including acoustic feeding devices and underwater fish-attracting lights. When feeding the target cobia fry batches in the target transport cages through the adaptive feeding system, the feeding images of the target cobia fry batches are acquired in real time through an underwater camera, labeled as target feeding images, and transmitted to the control terminal for image preprocessing to obtain preprocessed target feeding images. By analyzing and preprocessing the target feeding image, a specific region is delineated, and the rate of change of pixel values ​​in the specific region is analyzed. Combined with the amount of feed fed by the adaptive feeding system in the target transport cage, the change in feed in the target transport cage during the transfer analysis time is calculated, thereby generating the real-time feeding rate of the target cobia fry batch. Using the feeding as the starting point, the real-time feeding rate of the target cobia fry batches is analyzed, the feeding rate analysis time point is predetermined, and the danger threshold and standard threshold of the real-time feeding rate are predetermined. If the real-time feeding rate is greater than the danger threshold during the feeding rate analysis time point, it is determined that the target cobia fry batch is in a strong feeding stage. If the real-time feeding rate is less than the danger threshold but not less than the standard threshold, it is determined that the target cobia fry batch is in a stable feeding stage. If the real-time feeding rate is less than the standard threshold, it is determined that the target cobia fry batch is in a poor appetite stage. Based on the different feeding stages of the target cobia fry batches, a multimodal feeding strategy was formulated for the target cobia fry batches in combination with feeding induction equipment.

[0010] Furthermore, in a preferred embodiment of the present invention, the formulation of a multimodal feeding strategy for the target cobia fry batches based on different feeding stages and in conjunction with a feeding inducer device specifically includes: When the target cobia fry batch is in a stable feeding phase, there is no need to start the feeding equipment, and the current feeding amount of the adaptive feeding system should be kept unchanged. When the target cobia fry batch is in the intense feeding stage, the raw material ratio of the feed in the current adaptive feeding system is obtained and the raw material ratio is adjusted in real time until the real-time feeding rate of the target cobia fry batch in the target transport cage within a preset time is less than the danger threshold and not less than the standard threshold. When the target batch of cobia fry is in a stage of poor appetite, the operation of the feeding equipment is controlled to formulate a multimodal feeding strategy; The formulation of the multimodal feeding strategy involves obtaining a terminal system that controls all feeding devices and labeling it as a feeding device control system. Within the feeding device control system, different feeding modes of different feeding devices are freely combined to obtain different combinations of feeding device modes. Inside the target transport cage, the feeding device control system randomly outputs different combinations of feeding device modes and presets a feeding analysis time. During the feeding analysis time, the real-time feeding rate of the target cobia fry batch under the action of different combinations of feeding device modes is calculated. A combination of feeding device modes corresponding to a real-time feeding rate that is less than the danger threshold and not less than the standard threshold is used to generate and output a feeding strategy.

[0011] Furthermore, in a preferred embodiment of the present invention, the simultaneous implementation of intestinal health early warning and functional feed intervention for the target cobia fry batches during transportation via the target transport cage specifically includes: The underwater camera acquires real-time swimming video streams of target cobia fry batches and introduces a swimming posture anomaly detection model, which includes a target detection module, a key point extraction module, and a feature engineering module. The target monitoring module receives swimming video streams and performs video stream denoising, color correction, and contrast enhancement to obtain preprocessed swimming video streams. In the target monitoring module, the YOLO model is used to frame the locations of different target cobia fry and assign individual fry IDs. In the key point extraction module, key points of fish body features are detected for different target serpentine fry species within the frame, and feature vectors of fish body feature key points are generated. The feature engineering module collects the feature vectors of all fish body feature key points, constructs a normal distribution map, and calculates the mean and standard deviation of the feature vectors of different fish body feature key points to set the standard range of the feature vectors of fish body feature key points. Analyze the normal distribution map, mark the key points of fish body features whose feature vectors are not in the corresponding standard range as abnormal fish body feature points, and search the big data network for the intestinal health status stages corresponding to different abnormal fish body feature points of cobia fry. Search for abnormal intestinal health status stages within all intestinal health status stages. If the target cobia fry has abnormal characteristics that lead to an abnormal intestinal health status stage, then trigger an intestinal health warning in the target transport cage. A functional additive for regulating gut health is prepared. If a gut health warning is triggered in the target transport cage, the functional additive for regulating gut health is fed simultaneously with the feed through an adaptive feeding system until all target cobia fry show no abnormal characteristics.

[0012] The second aspect of this invention also provides an intelligent feeding and health management system suitable for land-sea relay aquaculture of cobia. The intelligent feeding and health management system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory consisting of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture. The memory contains a program for intelligent feeding and health management with a management engine. When the program is executed in parallel through a superscalar pipeline execution unit within the processor, the following steps are implemented: Feeding behavior analysis was conducted on cobia fry to generate a feeding activity score, and the cobia fry were screened and classified based on the score results; A transportation stress assessment was conducted on each batch of target cobia fry, and the stress resistance of the target cobia fry was strengthened based on the assessment results, while water quality parameters were controlled during transportation. Adaptive feeding training was conducted on cobia fry in the target transport cage, and image analysis was combined during the feeding training process to formulate multimodal feeding strategies. During the transportation of the target cobia fry through the target transport cages, intestinal health early warning and functional feed intervention were carried out simultaneously for each batch of target cobia fry.

[0013] This invention addresses the technical deficiencies in the prior art and offers the following beneficial effects: It achieves healthy land-sea relay farming of cobia fry through feeding screening and classification, transportation stress assessment, long-distance transportation technology optimization, water quality parameter control, and intelligent feeding training. Furthermore, it utilizes wind- and wave-resistant marine cages to provide a superior farming environment. This invention ensures the nutritional needs of the fry. Simultaneously, scientific farming management and timely disease observation and prevention guarantee the healthy and rapid growth of the fry. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0015] Figure 1 A flowchart is shown for an intelligent feeding and health management method suitable for land-sea relay farming of cobia; Figure 2 A flowchart illustrating the method for developing multimodal feeding strategies is shown. Figure 3 A program view of an intelligent feeding and health management system suitable for land-sea relay farming of cobia is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] Figure 1 A flowchart illustrating an intelligent feeding and health management method suitable for land-sea relay farming of cobia is shown, including the following steps: S102: Analyze the feeding behavior of cobia fry, generate a feeding enthusiasm score for cobia fry, and screen and classify cobia fry based on the score results; S104: Conduct a transport stress assessment on the target cobia fry batches, and enhance the stress resistance of the target cobia fry based on the assessment results, while controlling the water quality parameters during transport. S106: Adaptive feeding training of cobia fry was carried out in the target transport cage, and image analysis was combined during the feeding training process to formulate a multimodal feeding strategy; S108: During the transportation of target cobia fry through the target transport cage, intestinal health early warning and functional feed intervention are carried out simultaneously for the target cobia fry batches.

[0019] Furthermore, in a preferred embodiment of the present invention, the step of analyzing the feeding behavior of cobia fry to generate a feeding activity score for the cobia fry, and then screening and classifying the cobia fry based on the score results, specifically involves: Identify the current holding pond for cobia fry and designate it as the target holding pond, then install cameras in the target holding pond; Different batches of cobia fry were starved in the target holding tank. After starvation, live food was put into the target holding tank. At the same time, all the cameras were turned on to record real-time videos of the same batch of cobia fry feeding in the target holding tank, which were labeled as feeding behavior analysis videos. The feeding behavior analysis video is preprocessed, including video noise reduction and video interpolation optimization, to obtain a preprocessed feeding behavior analysis video. A trajectory tracking algorithm was introduced to perform trajectory tracking analysis on the same batch of cobia fry in pre-processed feeding behavior analysis videos, and to record the feeding behavior characteristics of the cobia fry. The feeding behavior characteristics of the cobia fry include reaction agility characteristics, feeding enthusiasm characteristics, competitive behavior characteristics, and feeding focus characteristics. A big data network is introduced to retrieve a feeding behavior characteristic scoring map. The feeding behavior characteristic scoring map records the scoring values ​​of different feeding behavior characteristics and presets a feeding enthusiasm scoring threshold. Among them, the feeding enthusiasm scoring threshold divides the fry of Cobra scad into an elimination group and a qualified group. The feeding behavior characteristics of cobia fry are imported into the feeding behavior characteristic scoring map for self-scoring. The feeding enthusiasm score of different batches of cobia fry is output. Combined with the feeding enthusiasm score threshold, the batches of cobia fry in the qualified group are marked as the target cobia fry batches.

[0020] It is important to note that selecting high-quality, healthy fry is fundamental to successful aquaculture. Before harvesting, the health status of the fry should be rigorously assessed, prioritizing those with excellent genetic quality, robust physique, no injuries or diseases, and uniform size. Cobia fry exhibit cannibalism; therefore, during selection and stocking, fry of uniform size must be strictly chosen for concentrated rearing to minimize losses due to cannibalism. Thus, temporary holding is necessary before land-sea relay aquaculture to screen for qualified batches of cobia fry. Feeding behavior analysis videos are used to record the feeding behavior characteristics of the cobia fry, including reaction agility, feeding enthusiasm, competitive behavior, and feeding focus. The reaction agility characteristic is calculated by determining the time required from baiting to the first fry entering the camera's designated "feeding interest zone"; the feeding enthusiasm characteristic is determined by counting the fry's mouth-opening actions near the bait and estimating the number of successful feedings; the competitive behavior characteristic is determined by using algorithms to identify rapid, angular "face-to-face" dashes and body contact between fry and counting the number of times each fry initiates such behavior; the feeding focus characteristic is determined by analyzing the fry's swimming path during feeding. Fish with high focus move more directly towards the bait with high linearity, while hesitant fish move more erratically. By analyzing different feeding characteristics, the enthusiasm of the fry can be differentiated to obtain the target cobia fry batch. Selecting the target cobia fry batch can fundamentally improve the overall quality and transport tolerance of the fry.

[0021] Furthermore, in a preferred embodiment of the present invention, the step of conducting a transport stress assessment on batches of target cobia fry and strengthening the stress resistance of the target cobia fry based on the assessment results, while controlling water quality parameters during transport, specifically includes: In the target holding tank, batches of target cobia fry were sampled and their blood oxygen saturation and behavioral stress were monitored. The behavioral stress monitoring tail subjected the sampled target cobia fry to stress treatment and recorded the changes in swimming speed of the samples after the stress treatment. The study analyzes the changes in swimming speed of samples after stress treatment, extracts the time when the swimming speed of samples drops to a preset threshold after stress treatment, calibrates it as the sample recovery time, and outputs the stress analysis status of the target cobia batch by combining the blood oxygen saturation of the samples. The planned transportation time is predetermined, and a transportation emergency vulnerability algorithm model is introduced. The transportation emergency vulnerability algorithm model combines the stress analysis status of the target cobia batch, the feeding enthusiasm score of the target cobia batch, and the planned transportation time to calculate the transportation emergency vulnerability index of the target cobia batch. By retrieving the nutritional fortification knowledge graph through big data networks, the transportation emergency vulnerability index of the target cobia batch is mapped to the nutritional fortification knowledge graph, and different levels of nutritional supplementation schemes corresponding to different transportation emergency vulnerability indices are marked in the nutritional fortification knowledge graph. In the target holding tank, based on the transportation emergency vulnerability index of the target cobia batch, a corresponding nutritional supplementation plan is output. In the process of outputting the corresponding nutritional supplementation plan, the feeding equipment is used to feed the fish in a quantitative manner that affects the supplementation plan. The system acquires the transport cages for the target batch of cobia and simultaneously monitors and controls the water quality parameters of the transport cages during the transport process.

[0022] It should be noted that the high loss rate of up to 30% during the transportation of cobia fry fully exposes the vulnerability of long-distance transportation. Therefore, it is necessary to strengthen and de-stress the fry before transshipment. Nutritional fortification and immunity enhancement can improve the fry's resistance to the stress of long-distance transportation. Before transportation, add an appropriate amount of multivitamins to the feed regularly and feed continuously for one week. Vitamin C, as an important antioxidant, can effectively reduce oxidative damage to the fish under stress, thereby enhancing immunity. At the same time, adding a certain amount of aquatic glucose to the tanks of the transport vehicle can effectively reduce stress during fry transportation, and a corresponding nutritional supplementation plan should be implemented. Secondly, blood oxygen saturation monitoring and behavioral stress monitoring are used to assess stress resistance. Blood oxygen saturation indirectly reflects the stability of the respiratory and circulatory systems. Behavioral stress testing involves applying a mild, sudden stress to the fish population, such as suddenly switching lights on and off for 5 seconds, or a brief, low-volume loud noise from the lights. The recovery time from being startled to returning to a calm, grouped state is recorded to test and output the stress analysis status of the target cobia batch. Subsequently, the transport vulnerability index (TVSI) of the target cobia batch is calculated. The calculation formula is as follows: , Where α, β, and γ are model weighting coefficients and are constants, H is blood oxygen saturation, T is sample recovery time, S is the feeding activity score of the target cobia batch, and D is the planned transportation duration. The output yields the transportation emergency vulnerability index, which corresponds to the corresponding nutrient supplementation plan.

[0023] Furthermore, in a preferred embodiment of the present invention, the step of acquiring the transport cages for the target batch of cobia and simultaneously monitoring and controlling the water quality parameters of the transport cages during the transport of the target cobia batch specifically includes: The transport cages for the target batch of cobia are identified and designated as target transport cages. After outputting the corresponding nutrient addition plan, the target batch of cobia is transferred from the target holding pond to the target transport cage. The target transport cage is equipped with a sensor network for real-time monitoring of water quality parameters. An adaptive feeding system is used to perform preliminary feeding on batches of target cobia fry transferred from the target holding tank to the target transport cage during the transfer process. The preliminary feeding process involves intelligently feeding the target transport cage according to the adaptive feeding system and controlling the amount of feed fed to be equal to the amount fed in the target holding tank. In the target transport cage, the water quality parameters of the target cobia batch are recorded immediately after transfer. The transfer analysis time is preset, and the changes in the water quality parameters of the target cobia batch during the transfer analysis time are recorded. Combined with calculations, the trend of water quality parameter changes in the target transport cage is generated. Intelligent water quality parameter adjustment equipment is introduced into the target transport cage. The trend of water quality parameter changes in the target transport cage is imported into the intelligent water quality parameter adjustment equipment, and a closed-loop intervention mechanism is set in the intelligent water quality parameter adjustment equipment. The closed-loop intervention mechanism involves setting an adjustment threshold for the water quality parameters that need to be intervened and adjusted based on the changing trend of the water quality parameters of the target transport cage, and controlling the intelligent water quality parameter adjustment device to control the water quality parameters when the changing trend of the water quality parameters of the target transport cage does not remain at the adjustment threshold, so that the water quality parameters are always maintained within the standard range.

[0024] It is important to note that water quality deterioration during transportation is the main cause of high mortality rates among fish fry, particularly the decrease in dissolved oxygen and the increase in ammonia nitrogen and nitrite. Therefore, precise control of the transportation water environment is crucial to ensuring the survival of the fish fry. Water quality parameters should be monitored and adjusted after feeding, as the excrement of the fry will pollute the water. Among the water quality parameters, cobia have extremely high requirements for dissolved oxygen. During transportation, continuous oxygenation with pure oxygen is necessary to ensure that the dissolved oxygen concentration in the water is maintained above 6.0 mg / L. For salinity, the suitable salinity for cobia is 10-35‰. For the transfer from ponds (which may be low-salinity or brackish water) to the sea (high-salinity seawater), drastic changes in salinity are a significant stressor leading to high mortality rates. Proper acclimation during stocking is essential to allow the fry to gradually adapt to the salinity changes. Ammonia nitrogen and nitrite are products of fish excrement and uneaten feed decomposition and are toxic to fish. Ammonia nitrogen should be strictly controlled below 0.1 mg / L. Regarding density, excessively high transport density will lead to rapid water quality deterioration, overcrowding, and significantly increased stress on the fry. The recommended density for transporting cobia fry (4-6 cm in length) in oxygen-rich water trucks is 5000-8000 fry / m³. For long-distance transport exceeding 20 hours, fresh seawater should be replaced en route to prevent water quality deterioration.

[0025] Furthermore, in a preferred embodiment of the present invention, the simultaneous implementation of intestinal health early warning and functional feed intervention for the target cobia fry batches during transportation via the target transport cage specifically includes: The underwater camera acquires real-time swimming video streams of target cobia fry batches and introduces a swimming posture anomaly detection model, which includes a target detection module, a key point extraction module, and a feature engineering module. The target monitoring module receives swimming video streams and performs video stream denoising, color correction, and contrast enhancement to obtain preprocessed swimming video streams. In the target monitoring module, the YOLO model is used to frame the locations of different target cobia fry and assign individual fry IDs. In the key point extraction module, key points of fish body features are detected for different target serpentine fry species within the frame, and feature vectors of fish body feature key points are generated. The feature engineering module collects the feature vectors of all fish body feature key points, constructs a normal distribution map, and calculates the mean and standard deviation of the feature vectors of different fish body feature key points to set the standard range of the feature vectors of fish body feature key points. Analyze the normal distribution map, mark the key points of fish body features whose feature vectors are not in the corresponding standard range as abnormal fish body feature points, and search the big data network for the intestinal health status stages corresponding to different abnormal fish body feature points of cobia fry. Search for abnormal intestinal health status stages within all intestinal health status stages. If the target cobia fry has abnormal characteristics that lead to an abnormal intestinal health status stage, then trigger an intestinal health warning in the target transport cage. A functional additive for regulating gut health is prepared. If a gut health warning is triggered in the target transport cage, the functional additive for regulating gut health is fed simultaneously with the feed through an adaptive feeding system until all target cobia fry show no abnormal characteristics.

[0026] It should be noted that during the high-stress period of transportation and training feeding, fry are prone to intestinal problems. Therefore, a swimming posture abnormality detection model is used to analyze the swimming characteristics of each fry to determine if intestinal problems are present. Intestinal health is assessed by analyzing key features, including body posture angles, swimming trajectory curvature, and tail wagging frequency and amplitude. Individual fry ID assignment aims to locate problematic fry, and normal distribution analysis is used to identify problematic feature points, enabling random intestinal monitoring and early warning. Enteritis is a typical and common disease in cobia farming, causing significant harm. Regular medicated feeding is an important preventative measure, typically once every two weeks for 3-5 consecutive days. Adding traditional Chinese medicine ingredients such as allicin and Sanhuang powder to the feed can further enhance the efficacy of the medication and improve the fry's disease resistance. Observe the fry's activity daily; if abnormal swimming is observed, take timely measures. For surface problems, fry can be netted and given a medicated bath. If an unknown disease is discovered, seek immediate diagnosis and treatment from a professional technician. Medicated feed is a combination of functional additives for regulating intestinal health and feed.

[0027] Figure 2 A flowchart illustrating a method for developing a multimodal feeding strategy is shown, including the following steps: S202: Adaptive feeding training of cobia fry was carried out in the target transport cage, and image analysis was combined during the feeding training process to formulate a multimodal feeding strategy; S204: Based on the different feeding stages of the target cobia fry batches, a multimodal feeding strategy was developed for the target cobia fry batches in combination with feeding equipment.

[0028] Furthermore, in a preferred embodiment of the present invention, the adaptive feeding training of cobia fry in the target transport cage, and the combination of image analysis during the feeding training process to formulate a multimodal feeding strategy, specifically includes: In the target transport cage, underwater cameras and feeding devices are installed, including acoustic feeding devices and underwater fish-attracting lights. When feeding the target cobia fry batches in the target transport cages through the adaptive feeding system, the feeding images of the target cobia fry batches are acquired in real time through an underwater camera, labeled as target feeding images, and transmitted to the control terminal for image preprocessing to obtain preprocessed target feeding images. By analyzing and preprocessing the target feeding image, a specific region is delineated, and the rate of change of pixel values ​​in the specific region is analyzed. Combined with the amount of feed fed by the adaptive feeding system in the target transport cage, the change in feed in the target transport cage during the transfer analysis time is calculated, thereby generating the real-time feeding rate of the target cobia fry batch. Using the feeding as the starting point, the real-time feeding rate of the target cobia fry batches is analyzed, the feeding rate analysis time point is predetermined, and the danger threshold and standard threshold of the real-time feeding rate are predetermined. If the real-time feeding rate is greater than the danger threshold during the feeding rate analysis time point, it is determined that the target cobia fry batch is in a strong feeding stage. If the real-time feeding rate is less than the danger threshold but not less than the standard threshold, it is determined that the target cobia fry batch is in a stable feeding stage. If the real-time feeding rate is less than the standard threshold, it is determined that the target cobia fry batch is in a poor appetite stage. Based on the different feeding stages of the target cobia fry batches, a multimodal feeding strategy was formulated for the target cobia fry batches in combination with feeding induction equipment.

[0029] It should be noted that underwater cameras and feeding devices are deployed to acquire videos of the fish fry swimming and feeding. During the process, the feeding rate is (total feed - change in feed volume within the target transport cage) / total feed. Different feeding rates require different feeding decisions. If the real-time feeding rate is greater than the danger threshold, the target cobia fry batch is judged to be in a strong feeding phase. If the real-time feeding rate is less than the danger threshold but not less than the standard threshold, the target cobia fry batch is judged to be in a stable feeding phase. If the real-time feeding rate is less than the standard threshold, the target cobia fry batch is judged to be in a poor appetite phase. Different feeding phases correspond to different feed acclimation strategies in the cages.

[0030] Furthermore, in a preferred embodiment of the present invention, the formulation of a multimodal feeding strategy for the target cobia fry batches based on different feeding stages and in conjunction with a feeding inducer device specifically includes: When the target cobia fry batch is in a stable feeding phase, there is no need to start the feeding equipment, and the current feeding amount of the adaptive feeding system should be kept unchanged. When the target cobia fry batch is in the intense feeding stage, the raw material ratio of the feed in the current adaptive feeding system is obtained and the raw material ratio is adjusted in real time until the real-time feeding rate of the target cobia fry batch in the target transport cage within a preset time is less than the danger threshold and not less than the standard threshold. When the target batch of cobia fry is in a stage of poor appetite, the operation of the feeding equipment is controlled to formulate a multimodal feeding strategy; The formulation of the multimodal feeding strategy involves obtaining a terminal system that controls all feeding devices and labeling it as a feeding device control system. Within the feeding device control system, different feeding modes of different feeding devices are freely combined to obtain different combinations of feeding device modes. Inside the target transport cage, the feeding device control system randomly outputs different combinations of feeding device modes and presets a feeding analysis time. During the feeding analysis time, the real-time feeding rate of the target cobia fry batch under the action of different combinations of feeding device modes is calculated. A combination of feeding device modes corresponding to a real-time feeding rate that is less than the danger threshold and not less than the standard threshold is used to generate and output a feeding strategy.

[0031] It should be noted that after the fry are released into the net cages, strict feeding acclimatization is required for the first week. Initially, a small amount of fish paste mixed with frozen brine shrimp can be fed 6-8 times a day to allow all fry to adapt to the new feeding environment. If they are in a strong feeding phase, gradually transition to a small amount of fish paste mixed with eel meal, adjusting the feed ratio to allow the fish to adapt. When the target cobia fry batch is showing signs of poor appetite, it is necessary to induce the fish to feed. Simultaneously, acoustic feed attractants and underwater fish lights should be turned on, and different combinations of feeding modes should be set to judge changes in feeding rate, such as "sound only," "light only," "sound and light combination," and "alternating flashing." The combination of feeding mode corresponding to a real-time feeding rate below the danger threshold but not below the standard threshold should be output to generate and output the feeding strategy. Furthermore, implementing a strategy of feeding small amounts frequently during the fry stage to ensure sufficient food supply is crucial for ensuring all fry are well-fed, grow uniformly, and reduce cannibalism, thereby effectively improving the survival rate.

[0032] like Figure 3As shown, the second aspect of the present invention also provides an intelligent training and health management system suitable for land-sea relay farming of cobia. The intelligent training and health management system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory consisting of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture. The memory contains a program for intelligent training and health management with a management engine. When the program is executed in parallel through a superscalar pipeline execution unit within the processor, the following steps are implemented: Feeding behavior analysis was conducted on cobia fry to generate a feeding activity score, and the cobia fry were screened and classified based on the score results; A transportation stress assessment was conducted on each batch of target cobia fry, and the stress resistance of the target cobia fry was strengthened based on the assessment results, while water quality parameters were controlled during transportation. Adaptive feeding training was conducted on cobia fry in the target transport cage, and image analysis was combined during the feeding training process to formulate multimodal feeding strategies. During the transportation of the target cobia fry through the target transport cages, intestinal health early warning and functional feed intervention were carried out simultaneously for each batch of target cobia fry.

[0033] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart feeding and health management method applicable to land-sea relay aquaculture of cobia, characterized in that, Includes the following steps: Feeding behavior analysis was conducted on cobia fry to generate a feeding activity score, and the cobia fry were screened and classified based on the score results; A transportation stress assessment was conducted on each batch of target cobia fry, and the stress resistance of the target cobia fry was strengthened based on the assessment results, while water quality parameters were controlled during transportation. Adaptive feeding training was conducted on cobia fry in the target transport cage, and image analysis was combined during the feeding training process to formulate multimodal feeding strategies. During the transportation of the target cobia fry through the target transport cages, intestinal health early warning and functional feed intervention were carried out simultaneously for each batch of target cobia fry.

2. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 1, characterized in that, The feeding behavior analysis of cobia fry was conducted to generate a feeding activity score for the cobia fry. Based on the score results, the cobia fry were screened and classified. Specifically: Identify the current holding pond for cobia fry and designate it as the target holding pond, then install cameras in the target holding pond; Different batches of cobia fry were starved in the target holding tank. After starvation, live food was put into the target holding tank. At the same time, all the cameras were turned on to record real-time videos of the same batch of cobia fry feeding in the target holding tank, which were labeled as feeding behavior analysis videos. The feeding behavior analysis video is preprocessed, including video noise reduction and video interpolation optimization, to obtain a preprocessed feeding behavior analysis video. A trajectory tracking algorithm was introduced to perform trajectory tracking analysis on the same batch of cobia fry in pre-processed feeding behavior analysis videos, and to record the feeding behavior characteristics of the cobia fry. The feeding behavior characteristics of the cobia fry include reaction agility characteristics, feeding enthusiasm characteristics, competitive behavior characteristics, and feeding focus characteristics. A big data network is introduced to retrieve a feeding behavior characteristic scoring map. The feeding behavior characteristic scoring map records the scoring values ​​of different feeding behavior characteristics and presets a feeding enthusiasm scoring threshold. Among them, the feeding enthusiasm scoring threshold divides the fry of Cobra scad into an elimination group and a qualified group. The feeding behavior characteristics of cobia fry are imported into the feeding behavior characteristic scoring map for self-scoring. The feeding enthusiasm score of different batches of cobia fry is output. Combined with the feeding enthusiasm score threshold, the batches of cobia fry in the qualified group are marked as the target cobia fry batches.

3. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 1, characterized in that, The process involves conducting a transport stress assessment on batches of target cobia fry, enhancing their stress resistance based on the assessment results, and controlling water quality parameters during transport. Specifically: In the target holding tank, batches of target cobia fry were sampled and their blood oxygen saturation and behavioral stress were monitored. The behavioral stress monitoring tail subjected the sampled target cobia fry to stress treatment and recorded the changes in swimming speed of the samples after the stress treatment. The study analyzes the changes in swimming speed of samples after stress treatment, extracts the time when the swimming speed of samples drops to a preset threshold after stress treatment, calibrates it as the sample recovery time, and outputs the stress analysis status of the target cobia batch by combining the blood oxygen saturation of the samples. The planned transportation time is predetermined, and a transportation emergency vulnerability algorithm model is introduced. The transportation emergency vulnerability algorithm model combines the stress analysis status of the target cobia batch, the feeding enthusiasm score of the target cobia batch, and the planned transportation time to calculate the transportation emergency vulnerability index of the target cobia batch. By retrieving the nutritional fortification knowledge graph through big data networks, the transportation emergency vulnerability index of the target cobia batch is mapped to the nutritional fortification knowledge graph, and different levels of nutritional supplementation schemes corresponding to different transportation emergency vulnerability indices are marked in the nutritional fortification knowledge graph. In the target holding tank, based on the transportation emergency vulnerability index of the target cobia batch, a corresponding nutritional supplementation plan is output. In the process of outputting the corresponding nutritional supplementation plan, the feeding equipment is used to feed the fish in a quantitative manner that affects the supplementation plan. The system acquires the transport cages for the target batch of cobia and simultaneously monitors and controls the water quality parameters of the transport cages during the transport process.

4. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 3, characterized in that, The process of acquiring the transport cages for the target batch of cobia, and simultaneously monitoring and controlling the water quality parameters of the transport cages during the transport of the target cobia batch, specifically involves: The transport cages for the target batch of cobia are identified and designated as target transport cages. After outputting the corresponding nutrient addition plan, the target batch of cobia is transferred from the target holding pond to the target transport cage. The target transport cage is equipped with a sensor network for real-time monitoring of water quality parameters. An adaptive feeding system is used to perform preliminary feeding on batches of target cobia fry transferred from the target holding tank to the target transport cage during the transfer process. The preliminary feeding process involves intelligently feeding the target transport cage according to the adaptive feeding system and controlling the amount of feed fed to be equal to the amount fed in the target holding tank. In the target transport cage, the water quality parameters of the target cobia batch are recorded immediately after transfer. The transfer analysis time is preset, and the changes in the water quality parameters of the target cobia batch during the transfer analysis time are recorded. Combined with calculations, the trend of water quality parameter changes in the target transport cage is generated. Intelligent water quality parameter adjustment equipment is introduced into the target transport cage. The trend of water quality parameter changes in the target transport cage is imported into the intelligent water quality parameter adjustment equipment, and a closed-loop intervention mechanism is set in the intelligent water quality parameter adjustment equipment. The closed-loop intervention mechanism involves setting an adjustment threshold for the water quality parameters that need to be intervened and adjusted based on the changing trend of the water quality parameters of the target transport cage, and controlling the intelligent water quality parameter adjustment device to control the water quality parameters when the changing trend of the water quality parameters of the target transport cage does not remain at the adjustment threshold, so that the water quality parameters are always maintained within the standard range.

5. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 1, characterized in that, The process of adaptive feeding training for cobia fry in the target transport cage, combined with image analysis during the training process, is used to formulate a multimodal feeding strategy. Specifically: In the target transport cage, underwater cameras and feeding devices are installed, including acoustic feeding devices and underwater fish-attracting lights. When feeding the target cobia fry batches in the target transport cages through the adaptive feeding system, the feeding images of the target cobia fry batches are acquired in real time through an underwater camera, labeled as target feeding images, and transmitted to the control terminal for image preprocessing to obtain preprocessed target feeding images. By analyzing and preprocessing the target feeding image, a specific region is delineated, and the rate of change of pixel values ​​in the specific region is analyzed. Combined with the amount of feed fed by the adaptive feeding system in the target transport cage, the change in feed in the target transport cage during the transfer analysis time is calculated, thereby generating the real-time feeding rate of the target cobia fry batch. Using the feeding as the starting point, the real-time feeding rate of the target cobia fry batches is analyzed, the feeding rate analysis time point is predetermined, and the danger threshold and standard threshold of the real-time feeding rate are predetermined. If the real-time feeding rate is greater than the danger threshold during the feeding rate analysis time point, it is determined that the target cobia fry batch is in a strong feeding stage. If the real-time feeding rate is less than the danger threshold but not less than the standard threshold, it is determined that the target cobia fry batch is in a stable feeding stage. If the real-time feeding rate is less than the standard threshold, it is determined that the target cobia fry batch is in a poor appetite stage. Based on the different feeding stages of the target cobia fry batches, a multimodal feeding strategy was formulated for the target cobia fry batches in combination with feeding induction equipment.

6. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 5, characterized in that, The aforementioned multimodal feeding strategy for target cobia fry batches, based on different feeding stages and combined with feeding-inducing equipment, is as follows: When the target cobia fry batch is in a stable feeding phase, there is no need to start the feeding equipment, and the current feeding amount of the adaptive feeding system should be kept unchanged. When the target cobia fry batch is in the intense feeding stage, the raw material ratio of the feed in the current adaptive feeding system is obtained and the raw material ratio is adjusted in real time until the real-time feeding rate of the target cobia fry batch in the target transport cage within a preset time is less than the danger threshold and not less than the standard threshold. When the target batch of cobia fry is in a stage of poor appetite, the operation of the feeding equipment is controlled to formulate a multimodal feeding strategy; The formulation of the multimodal feeding strategy involves obtaining a terminal system that controls all feeding devices and labeling it as a feeding device control system. Within the feeding device control system, different feeding modes of different feeding devices are freely combined to obtain different combinations of feeding device modes. Inside the target transport cage, the feeding device control system randomly outputs different combinations of feeding device modes and presets a feeding analysis time. During the feeding analysis time, the real-time feeding rate of the target cobia fry batch under the action of different combinations of feeding device modes is calculated. A combination of feeding device modes corresponding to a real-time feeding rate that is less than the danger threshold and not less than the standard threshold is used to generate and output a feeding strategy.

7. The intelligent feeding and health management method for cobia in land-sea relay aquaculture as described in claim 1, characterized in that, During the transportation of the target cobia fry in the target transport cages, intestinal health early warning and functional feed intervention are simultaneously implemented for each batch. Specifically: The underwater camera acquires real-time swimming video streams of target cobia fry batches and introduces a swimming posture anomaly detection model, which includes a target detection module, a key point extraction module, and a feature engineering module. The target monitoring module receives swimming video streams and performs video stream denoising, color correction, and contrast enhancement to obtain preprocessed swimming video streams. In the target monitoring module, the YOLO model is used to frame the locations of different target cobia fry and assign individual fry IDs. In the key point extraction module, key points of fish body features are detected for different target serpentine fry species within the frame, and feature vectors of fish body feature key points are generated. The feature engineering module collects the feature vectors of all fish body feature key points, constructs a normal distribution map, and calculates the mean and standard deviation of the feature vectors of different fish body feature key points to set the standard range of the feature vectors of fish body feature key points. Analyze the normal distribution map, mark the key points of fish body features whose feature vectors are not in the corresponding standard range as abnormal fish body feature points, and search the big data network for the intestinal health status stages corresponding to different abnormal fish body feature points of cobia fry. Search for abnormal intestinal health status stages within all intestinal health status stages. If the target cobia fry has abnormal characteristics that lead to an abnormal intestinal health status stage, then trigger an intestinal health warning in the target transport cage. A functional additive for regulating gut health is prepared. If a gut health warning is triggered in the target transport cage, the functional additive for regulating gut health is fed simultaneously with the feed through an adaptive feeding system until all target cobia fry show no abnormal characteristics.

8. An intelligent feeding and health management system suitable for land-sea relay aquaculture of cobia, characterized in that, The intelligent bait training and health management system integrates a high-performance computing architecture and a data storage module, including a non-volatile memory consisting of a DDR4 RDIMM memory module with ECC verification and an NVMe solid-state storage array using 3D NAND flash memory, and a multi-core processor based on the Zen4 microarchitecture; the memory contains an intelligent bait training and health management method program with a management engine, and when the program is decoded and executed in parallel by the superscalar pipeline execution unit in the processor, the intelligent bait training and health management steps as described in any one of claims 1-7 are implemented.

Images