A feed regulation method and system based on the growth state of fry

Abstract

The present application relates to the field of intelligent management of aquaculture, and discloses a feed regulation method and system based on the growth state of fry, which comprises the following steps: acquiring growth and activity state parameters; establishing a multi-dimensional data information graph; generating an initial feeding scheme; outputting an optimal feeding scheme; performing closed-loop iterative feeding scheme; and performing monitoring system calibration and maintenance; the system comprises a parameter acquisition module, a data processing module, a feeding scheme generation module, a feeding scheme execution optimization module, and a data feedback dynamic optimization module; the present application realizes accurate matching of feed types, feeding amounts, and feeding times, and achieves the purposes of saving feed costs, promoting uniform growth of fry, and reducing the risk of aquaculture.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology for aquaculture, and in particular to a method and system for feed regulation based on the growth status of fish fry. Background Technology

[0002] Fish fry rearing is a crucial link in the aquaculture industry chain, and its growth quality directly determines the survival rate and economic benefits of subsequent adult fish farming. Traditional fish fry feeding methods mainly rely on the experience and judgment of farmers, which has the following core problems:

[0003] 1. Inefficient monitoring methods: There is a lack of real-time quantitative monitoring of the growth status (body length, weight, plumpness, etc.) and activity status (feeding activity, swimming frequency, etc.) of fish fry, making it impossible to accurately capture the growth dynamics of fish fry;

[0004] 2. Fixed feeding plan: Parameters such as feed type, feeding amount, and feeding time are mostly fixed values ​​and are not dynamically adjusted according to the real-time growth and activity characteristics of fish fry and environmental changes, resulting in feed waste or insufficient nutrient supply.

[0005] 3. Lack of data correlation: No quantitative correlation has been established between growth status, activity status and feeding effect, making it difficult to optimize feeding strategy through data-driven approaches;

[0006] 4. Poor environmental adaptability: The impact of environmental parameters such as water temperature, dissolved oxygen, and ammonia nitrogen concentration on the feeding and growth of fish fry was not fully considered, resulting in a feeding program with weak resistance to interference.

[0007] 5. Lack of closed-loop optimization: There is no effective feedback mechanism after feeding, and the feeding plan cannot be iteratively optimized based on data such as the amount of food consumed by the fry and changes in growth. In the long term, problems such as uneven growth and frequent diseases are likely to occur.

[0008] In existing technologies, some aquaculture feeding equipment only achieves timed and quantitative feeding functions, without combining dynamic monitoring and intelligent control of fish fry growth status; a few solutions using sensor monitoring suffer from drawbacks such as single monitoring parameters, simple data processing, and low model prediction accuracy, making it difficult to meet the precision feeding requirements of intensive aquaculture. Therefore, there is an urgent need to develop a feed regulation method and system based on multi-parameter real-time monitoring, in-depth data analysis, and intelligent model control to solve the problem of the extensive nature of traditional feeding methods. Summary of the Invention

[0009] The present invention aims to provide a method and system for feed regulation based on the growth status of fish fry, so as to achieve precise matching of feed type, feeding amount and feeding time, thereby saving feed costs, promoting uniform growth of fish fry and reducing aquaculture risks.

[0010] To achieve the above objectives, the present invention provides the following method:

[0011] This invention provides a method for regulating feed based on the growth status of fish fry:

[0012] S1: A monitoring system consisting of an image acquisition device and a biosensor array is used to monitor fish fry in aquaculture water in real time and obtain growth status monitoring parameters and activity status monitoring parameters.

[0013] S2. Based on the growth status monitoring parameters and activity status monitoring parameters collected in S1, establish multiple data information graphs, including a growth trend graph, an activity status distribution graph, and a growth-activity correlation graph, to present the correlation characteristics between fish fry growth and activity.

[0014] S3. Analyze the data information graph established in S2, extract the peak growth period, active feeding period and the correlation characteristics between the two, and formulate an initial feeding plan with clear feed type, single feeding amount, daily feeding frequency and time window in combination with the basic nutritional requirements of fish fry.

[0015] S4. Based on the initial feeding scheme as a sample, multiple environmental parameters such as water temperature, dissolved oxygen, and pH are incorporated. A feed regulation model is constructed using a gradient boosting decision tree algorithm. Historical fish fry growth activity data and corresponding feeding effect data are input to train the feed regulation model. A mapping relationship between multiple parameters and the feeding scheme is established, and the optimal feeding scheme, including the best feed type, precise feeding amount, and optimal feeding time, is output.

[0016] S5. Apply the optimal feeding scheme to actual aquaculture. The monitoring system collects data on the remaining amount of feed, growth and activity status of the fish fry after feeding in real time, and feeds it back to the feed control model. The feed control model is iteratively optimized at regular intervals to achieve dynamic optimization of the optimal feeding scheme.

[0017] Preferably, the image acquisition device includes an underwater high-definition camera and an infrared supplementary lighting module; the underwater high-definition camera adopts an IP68-level waterproof and dustproof sealed design, with a pixel count ≥ 8 million, a frame rate of 25-30fps, and a focal length of 8-12mm. A single unit covers a radius of 5-8 meters within the aquaculture water body, and is evenly distributed in three layers (top, middle, and bottom) within the aquaculture pond, with one unit per 100㎡ to ensure no blind spots; the infrared supplementary lighting module has a wavelength range of 850-940nm, and the illumination intensity is dynamically adjusted in real time based on feedback data from a water transparency sensor. When the transparency is ≤15cm, the illumination intensity... The system improves performance by 50%, effectively covering over 90% of the monitoring area. The biosensor array consists of body length and weight monitoring sensors, motion state sensors, feeding behavior sensors, and clustered sensors. It adopts a modular plug-and-play design, and the data acquisition frequency is synchronized with the image acquisition device, performing synchronization once every 0.5-2 hours. The body length and weight sensors achieve non-contact detection through laser ranging and image recognition fusion technology. The motion state sensors capture the instantaneous swimming trajectory of fish fry based on triaxial acceleration sensing technology. The static detection error of all sensors is ≤3%, and the dynamic detection error is ≤5%.

[0018] Preferably, the growth status monitoring parameters include body length, weight, average daily growth rate, body width, plumpness, and scale development integrity: the body width is the lateral distance at the widest point of the fry's body;

[0019] The formula for calculating the fullness is:

[0020] ;

[0021] in, For plumpness, The length of the fish fry, For the width of the fish fry, As the benchmark factor, and The product of these factors constitutes the phenotypic volume characteristic of the fish fry, which is then compared with the baseline factor. The ratio eliminates the influence of individual size differences on developmental assessment and realizes the quantification of "fatness / thinness";

[0022] The scale development integrity is assessed by statistically analyzing scale coverage and scale damage ratio using image recognition. Activity status monitoring parameters include swimming frequency, feeding activity, cluster distribution density, stress response frequency, diurnal activity difference coefficient, and feeding response time. The data acquisition interval for the growth status monitoring parameters and the activity status monitoring parameters is 0.5-2 hours, with 0.5-1 hours during the seedling stage and 1-2 hours during the growth stage. When the daily growth rate of body length exceeds 10% or the stress response frequency exceeds the limit three times consecutively, the data acquisition interval is shortened to 0.5 hours to ensure the timeliness of data acquisition during key growth stages.

[0023] Preferably, the core data sources are the growth status monitoring parameters (body length, weight, and daily growth rate) and the activity status monitoring parameters (swimming frequency, feeding activity, and cluster distribution density) collected in step S1. The coordinate axes and presentation logic of each chart are clearly defined: the growth trend chart uses time as the horizontal axis, marking nodes at 1-2 hour collection intervals, and simultaneously displays the values ​​of the three types of growth parameters on the vertical axis. Different colored lines are used to fit the data points, intuitively presenting the changing patterns of single parameters over time and the collaborative growth / fluctuation characteristics of multiple parameters. The activity status distribution chart uses the activity parameter type as the horizontal axis and the frequency proportion of the corresponding parameter on the vertical axis, using a bar chart with error bars to mark the mean, standard deviation, and peak range of each parameter. The growth-activity correlation chart uses a dual-vertical-axis scatter plot, with the horizontal axis representing feeding activity, the left vertical axis representing daily growth rate, and the right vertical axis representing weight growth rate. The positive / negative correlation between the two is quantified through the degree of clustering of data points and the slope of the fitted curve. The growth trend chart, activity status distribution chart, and growth-activity correlation chart all support parameter filtering and time period scaling.

[0024] Preferably, the basic nutritional requirements of the fish fry species include: clearly quantified indicators: protein 35%-55%, fat 8%-15%, vitamin A:vitamin D:vitamin E:vitamin C = 3:1:2:5, and minerals calcium ≥0.8% and phosphorus ≥0.6%. The initial feeding scheme is as follows: the ratio of feed pellet size to the average mouth diameter of the fish fry is 1:1.2-1:1.5; feeding uniformity is randomly detected by the monitoring system to ensure density difference ≤10%; and the duration of each feeding session is 15-30 minutes. All parameters are calibrated based on the characteristics of the fish fry species, growth stage, and data from at least three previous feeding and rearing batches.

[0025] Preferably, when constructing the feed regulation model, the environmental parameters further include controlling ammonia nitrogen concentration ≤0.2mg / L, nitrite concentration ≤0.1mg / L, water turbidity ≤20NTU, and water flow velocity in the middle layer of the water body to 0.1-0.3m / s; the growth status monitoring parameters and activity status monitoring parameters are synchronously collected using a dedicated water quality sensor, with the collection frequency consistent with the growth status monitoring parameters, data transmission delay ≤10 seconds, and the collected data is first normalized to map it to the [0,1] interval to remove extreme value interference; the gradient boosting decision tree algorithm's parameter dynamic adaptation is as follows: the initial learning rate is 0.05, which is reduced to 0.03 when the training set fitting error >5%, and the fitting error... When the percentage is <2%, the value is increased to 0.08; the decision tree depth is adjusted according to the growth stage of the fish fry, with 3-5 layers in the fry stage and 6-8 layers in the growth stage, the minimum number of samples in the leaf nodes is ≥10, and the regularization coefficient is 0.005 to suppress overfitting; the training steps of the feed regulation model are as follows: the historical fish fry feeding data is divided into training set and test set in a 7:3 ratio, the training set is cross-validated with 5 folds, the data in each fold is randomly divided, and the average value is taken after repeating 3 times to optimize the parameters. The prediction accuracy of the feed regulation model for the core parameters of the optimal feeding scheme is ≥92%, the prediction deviation of the average daily growth rate and the feeding activity is ≤8%, and the deviation of non-core parameters is ≤12%, providing stable algorithm support for the dynamic iterative optimization of S5.

[0026] Preferably, the optimal feeding scheme is composed of core parameters output by the feed regulation model: including the optimal feed type to meet the nutritional needs of the fry, the precise feeding amount calculated based on the weight percentage, the optimal feeding time based on the active feeding period, and the feeding method suitable for the density; the optimal feeding scheme is equipped with an automatic switching mechanism: when the stocking density exceeds 50 fish / m² or the daily growth rate of body length is >8%, the feeding amount is dynamically adjusted; the pause condition is that the feeding activity is <3 times / minute for 5 consecutive minutes or the remaining feed is more than 20%, and the emergency scenario is to quickly adjust according to the growth status monitoring parameters and the activity status monitoring parameters, and synchronously link the monitoring system of S5 to collect data on the remaining feed amount, growth and activity status changes of the fry after feeding in real time.

[0027] Preferably, the optimal feeding scheme is applied to actual aquaculture. The monitoring system collects real-time data on the remaining feed intake, growth, and activity status of the fish fry after feeding, and feeds this data back to the feed control model. The feed control model is periodically iterated and optimized to achieve dynamic optimization of the optimal feeding scheme. This includes: after feeding according to the optimal feeding scheme, collecting four types of feedback data through the monitoring system: remaining feed intake, changes in growth status, activity response, and water quality correlation data, while also incorporating fish fry mortality and disease incidence rates; and adjusting the feed control model's iteration cycle according to data wave... Dynamic adjustment: 24 hours / time during the seedling stage, 48-72 hours / time during the growth stage, shortened to 12 hours when parameter fluctuation exceeds 15%; The iterative optimization process adopts an incremental learning algorithm, with new data updated according to time weight, where 0.8 is given for the first 7 days, 0.7 for 8-30 days, and 0.6 for more than 30 days. Outliers are eliminated using the 3σ principle. If the prediction accuracy of the growth status monitoring parameters and the activity status monitoring parameters improves by ≥5% after iteration, the optimal feeding scheme is updated; otherwise, a data anomaly report is generated to investigate equipment or environmental problems, forming a closed loop of "feeding-monitoring-optimization".

[0028] Preferably, after step S5, the method further includes: to ensure the accuracy of the model input data and the optimization effect of S5, the monitoring system is calibrated and maintained every 7-15 days. When the water quality is turbid or the parameter fluctuation exceeds 20%, the calibration and maintenance time is shortened to 7 days. The calibration and maintenance includes: cleaning the lens of the underwater high-definition camera and calibrating it with a standard reference to ensure that the body length measurement error is ≤2%; calibrating the sensitivity of the biosensor array in a standard simulated environment, with a detection error of ≤3%; the signal strength of the data transmission module that transmits the growth status monitoring parameters and activity status monitoring parameters collected by the monitoring system to the computer must be ≥-70dBm, and the wireless transmission success rate must be ≥99%; establishing encrypted aquaculture records, recording calibration data, model iteration logs, feeding scheme adjustment records, and fry growth reports, and storing them in the cloud for ≥2 years to provide historical data support for the training of the feed regulation model and achieve continuous optimization of the optimal feeding scheme.

[0029] This invention discloses a feed regulation system based on the growth status of fish fry, characterized in that the system comprises:

[0030] Parameter acquisition module: The monitoring system is composed of an image acquisition device and a biosensor array to monitor fish fry in aquaculture water in real time and acquire growth status monitoring parameters and activity status monitoring parameters;

[0031] Data processing module: Based on the collected growth status monitoring parameters and activity status monitoring parameters, establish multiple data information graphs, including growth trend graph, activity status distribution graph, and growth-activity correlation graph, to present the correlation characteristics between fish fry growth and activity;

[0032] Feeding plan generation module: Analyze the established data information graph, extract the peak growth period, active feeding period and the correlation characteristics between the two, and formulate an initial feeding plan with clear feed type, single feeding amount, daily feeding frequency and time window in combination with the basic nutritional requirements of fish fry.

[0033] Feeding scheme execution optimization module: Based on the initial feeding scheme as the base sample, multiple environmental parameters such as water temperature, dissolved oxygen, and pH value are incorporated. A feed regulation model is constructed using a gradient boosting decision tree algorithm. The feed regulation model is trained by inputting historical fish fry growth activity data and corresponding feeding effect data, establishing a mapping relationship between multiple parameters and the feeding scheme, and outputting the optimal feeding scheme including the best feed type, precise feeding amount, and optimal feeding time.

[0034] Data feedback dynamic optimization module: The optimal feeding scheme is applied to actual aquaculture. The monitoring system collects data on the remaining amount of feed, growth and activity status of fish fry after feeding in real time, and feeds it back to the feed control model. The feed control model is iteratively optimized at regular intervals to achieve dynamic optimization of the optimal feeding scheme.

[0035] The beneficial effects of this invention are as follows: By employing a monitoring method that integrates image acquisition and a biosensor array, this invention achieves non-contact, multi-dimensional, and blind-spot-free monitoring of growth and activity parameters, with a data acquisition error of ≤5%, solving the problem of inefficient traditional monitoring. Through multi-dimensional data chart analysis of the correlation between growth and activity, and combined with a gradient boosting decision tree algorithm to construct a control model, it achieves precise matching of feed type, feeding amount, and feeding time, with a core parameter prediction accuracy of ≥92%. It establishes a closed-loop mechanism of "feeding-monitoring-optimization," adjusting the feeding plan in real time according to changes in fry growth, activity response, and environmental fluctuations, adapting to different growth stages and aquaculture environments, and improving the adaptability of the plan. It reduces feed waste (residual feed percentage ≤5%), promotes uniform fry growth (body length difference ≤10%), reduces disease incidence (≤3%), saves labor costs, and improves survival rate and economic benefits. The monitoring system adopts a modular design, supports calibration, maintenance, and troubleshooting, with a data transmission success rate of ≥99%, and aquaculture records are stored in the cloud for ≥2 years, providing data support for long-term aquaculture. Attached Figure Description

[0036] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0037] Figure 1A schematic flowchart of a feed regulation method based on the growth status of fish fry provided in an embodiment of the present invention;

[0038] Figure 2 This is a schematic diagram of a feed regulation system based on the growth status of fish fry, provided as an embodiment of the present invention. Detailed Implementation

[0039] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0041] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0042] In existing technologies, some aquaculture feeding equipment only achieves timed and quantitative feeding functions, without combining dynamic monitoring and intelligent control of fish fry growth status; a few solutions using sensor monitoring suffer from drawbacks such as single monitoring parameters, simple data processing, and low model prediction accuracy, making it difficult to meet the precision feeding requirements of intensive aquaculture. Therefore, there is an urgent need to develop a feed regulation method and system based on multi-parameter real-time monitoring, in-depth data analysis, and intelligent model control to solve the problem of the extensive nature of traditional feeding methods.

[0043] The present invention aims to provide a method and system for feed regulation based on the growth status of fish fry, so as to achieve precise matching of feed type, feeding amount and feeding time, thereby saving feed costs, promoting uniform growth of fish fry and reducing aquaculture risks.

[0044] like Figure 1 As shown in the figure, a specific embodiment of the present invention provides a method for feed regulation based on the growth status of fish fry, including the following steps:

[0045] 1. Real-time monitoring: Acquire growth and activity status parameters

[0046] A monitoring system composed of image acquisition devices and biosensor arrays is used to conduct comprehensive, blind-spot-free real-time monitoring of fish fry in aquaculture water. The specific implementation method is as follows:

[0047] Image acquisition device deployment: Includes an underwater high-definition camera and an infrared supplementary lighting module. The underwater high-definition camera adopts an IP68-level waterproof and dustproof sealed design, with a pixel count ≥ 8 million, a frame rate of 25-30fps, and a focal length of 8-12mm. A single unit covers a radius of 5-8 meters in the aquaculture water body. It is arranged in three layers within the aquaculture pond, with one unit deployed per 100㎡ to ensure no blind spots. The infrared supplementary lighting module has a wavelength range of 850-940nm, and the illumination intensity is dynamically adjusted in real time based on feedback data from the water transparency sensor: when the transparency is ≤15cm, the illumination intensity is automatically increased by 50%, ensuring image clarity in over 90% of the monitored area.

[0048] Biosensor array configuration: Composed of body length and weight monitoring sensors, motion status sensors, feeding behavior sensors, and clustered sensors, employing a modular plug-and-play design for easy maintenance and replacement. Functions and technical parameters of each sensor:

[0049] Body length and weight sensor: Through the fusion technology of laser ranging and image recognition, non-contact detection is achieved, avoiding stress to fish fry;

[0050] Motion state sensor: Based on three-axis accelerometer technology, it captures the instantaneous swimming trajectory of fish fry and quantifies the swimming frequency;

[0051] Feeding behavior sensor: monitors the frequency and duration of feeding actions in fish fry, and calculates feeding activity level;

[0052] Cluster distribution sensor: detects the spatial distribution density of fish fry and determines the cluster status.

[0053] All sensors have a static detection error of ≤3% and a dynamic detection error of ≤5%, and the data acquisition frequency is synchronized with the image acquisition device.

[0054] Monitoring parameters and data acquisition frequency:

[0055] Growth status monitoring parameters include body length, weight, average daily growth rate, body width (lateral distance at the widest point of the trunk), condition (calculation formula: condition = (weight / body length³) × 100), and scale development integrity (scale coverage and damage ratio are statistically analyzed through image recognition).

[0056] Activity status monitoring parameters include swimming frequency, feeding activity, cluster distribution density, stress response frequency, diurnal activity difference coefficient, and feeding response time;

[0057] The sampling frequency is dynamically adjusted: during the fry stage (body length < 3cm), the sampling frequency is 0.5-1 hour / time, and during the growth stage (body length ≥ 3cm), the sampling frequency is 1-2 hours / time. When the daily growth rate of body length exceeds 10% or the stress response frequency exceeds the limit for 3 consecutive times (the stress response frequency threshold is set according to the fish fry species, such as the threshold for California bass fry being 5 times / hour), the sampling interval is automatically shortened to 0.5 hours to ensure that no data from key growth stages is missed.

[0058] 2. Data Processing: Establishing Multi-Dimensional Data Infographics

[0059] Based on the core parameters collected in step 1 (body length, weight, average daily growth rate, swimming frequency, feeding activity, and school density), three types of data information charts are established using data visualization technology to intuitively present the correlation characteristics between fish fry growth and activity:

[0060] Growth trend chart: The horizontal axis represents time (with nodes marked at 1-2 hour collection intervals), and the vertical axis simultaneously displays the values ​​of three parameters: body length, weight, and daily average growth rate. Red lines represent body length, blue lines represent weight, and green lines represent daily average growth rate. By fitting data points with lines, the chart clearly presents the changing patterns of single parameters over time and the synergistic growth / fluctuation characteristics of multiple parameters.

[0061] Activity status distribution chart: The horizontal axis represents the type of activity parameter (swimming frequency, feeding activity, cluster distribution density, etc.), and the vertical axis represents the frequency percentage of the corresponding parameter; a bar chart with error bars is used to mark the mean, standard deviation and peak range of each parameter (e.g., the peak range of feeding activity is 8-12 times / minute) to facilitate quick identification of abnormal activity status.

[0062] Growth-activity correlation plot: A scatter plot with two vertical axes is used. The horizontal axis represents feeding activity, the left vertical axis represents the average daily growth rate, and the right vertical axis represents the weight gain rate. The positive correlation between the two is quantified by the degree of clustering of data points (clustering coefficient ≥ 0.7 indicates strong correlation) and the slope of the fitted curve (fitted curve slope > 0.5 indicates strong positive correlation).

[0063] All three types of charts support parameter filtering (such as showing only the relationship between body length and feeding activity) and time period scaling (such as focusing on data from the last 24 hours), which facilitates targeted analysis by farmers.

[0064] 3. Plan Formulation: Generate an initial feeding plan.

[0065] By analyzing the data information graph established in step 2 and combining it with the basic nutritional requirements of the fish fry species, an initial feeding plan is formulated:

[0066] Key feature extraction: Extract peak growth periods (e.g., the period with the highest daily growth rate from 8:00 to 12:00) from the growth trend map; extract active feeding periods (e.g., peak feeding activity from 10:00 to 11:00 and 16:00 to 17:00) from the activity status distribution map; and extract correlation features between the two (e.g., a positive correlation coefficient of 0.85 between feeding activity and weight gain rate) from the growth-activity correlation map.

[0067] Basic nutritional requirements: clearly quantified indicators: protein 35%-55% (50%-55% in seedling stage, 35%-40% in growth stage), fat 8%-15%, vitamin ratio (A:D:E:C=3:1:2:5), minerals (calcium ≥0.8%, phosphorus ≥0.6%).

[0068] Initial feeding plan parameters:

[0069] Feed type: Select compound feed with corresponding proportions according to nutritional needs (such as high-protein feed for seedlings and balanced nutrition feed for growth period).

[0070] Particle size: The ratio of particle size to the average mouth diameter of the fish fry should be 1:1.2-1:1.5 (e.g., for fish fry with an average mouth diameter of 2mm, choose feed with a particle size of 2.4-3mm).

[0071] Single feeding amount: 3%-5% of the fry's body weight (calibrated based on breed characteristics);

[0072] Daily feeding frequency: determined according to the active feeding period, usually 2-4 times (e.g., if the active feeding period is 10:00 and 16:00, set to feed 2 times a day);

[0073] Time window: Match the peak feeding times (e.g., 10:00-10:30, 16:00-16:30);

[0074] Feeding uniformity: Random testing through a monitoring system ensures that the feed density difference within the aquaculture pond is ≤10%;

[0075] Feeding time per feeding: 15-30 minutes (15 minutes for seedlings and 30 minutes for adult plants).

[0076] All parameters are calibrated based on the characteristics of the fish fry species, growth stage, and data from at least three previous feeding and rearing cycles to ensure the rationality of the initial plan.

[0077] 4. Model Building: Output the optimal feeding scheme

[0078] Using the initial feeding plan as the base sample, a feed regulation model is constructed by combining environmental parameters. This model is then trained and optimized using historical data to output the optimal feeding plan.

[0079] Environmental parameters included: water temperature (suitable range 20-28℃), dissolved oxygen (≥5mg / L), pH (7.0-8.5), ammonia nitrogen concentration (≤0.2mg / L), nitrite concentration (≤0.1mg / L), turbidity (≤20NTU), and water flow velocity (0.1-0.3m / s); a dedicated water quality sensor was used to collect data synchronously with growth / activity parameters, with the collection frequency consistent with the growth status monitoring parameters, and data transmission delay ≤10 seconds; after collection, the data was first normalized (mapped to the [0,1] interval), and extreme value interference (such as dissolved oxygen abnormal values ​​<3mg / L) was removed by the 3σ principle.

[0080] Feed regulation model construction: Gradient boosting decision tree (GBDT) algorithm with dynamic parameter adaptation.

[0081] Learning rate: Initial value 0.05, adjusted to 0.03 when the training set fitting error is >5%, and adjusted to 0.08 when the fitting error is <2%;

[0082] Decision tree depth: 3-5 layers in the seedling stage, 6-8 layers in the growth stage;

[0083] Regularization parameter: minimum number of samples per leaf node ≥ 10, regularization coefficient 0.005 (to suppress overfitting);

[0084] Model training process:

[0085] Data partitioning: Historical fish fry rearing data (including growth status, activity status, environmental parameters, feeding plan, and feeding effect) were divided into training set and test set in a 7:3 ratio;

[0086] Cross-validation: The training set is cross-validated with a 5-fold split, where the data in each fold is randomly divided and the process is repeated 3 times, and the average value is taken to optimize the parameters.

[0087] Model accuracy requirements: Prediction accuracy of core parameters (optimal feed type, precise feeding amount, optimal feeding time) ≥92%, prediction deviation of daily growth rate and feeding activity ≤8%, and deviation of non-core parameters (such as feeding duration) ≤12%;

[0088] Optimal feeding plan output:

[0089] Key parameters: the optimal feed type to meet the nutritional needs of fish fry, the precise feeding amount calculated as a percentage of body weight (e.g., 5% ± 0.2% of body weight), the optimal feeding time based on the period of active feeding (accurate to the minute), and the feeding method appropriate to the density (e.g., multi-point feeding in high-density aquaculture).

[0090] Automatic switching mechanism: When the stocking density exceeds 50 fish / m² or the daily growth rate of body length is greater than 8%, the feeding amount will be dynamically adjusted (e.g., the feeding amount will be increased by 10%).

[0091] Suspension conditions: If the feeding activity level is less than 3 times / minute for 5 consecutive minutes or more than 20% of the feed remains, feeding should be suspended to avoid feed waste.

[0092] 5. Dynamic optimization: Closed-loop iterative feeding scheme

[0093] The optimal feeding plan is applied to actual aquaculture, and the plan is dynamically iteratively optimized by using data feedback from the monitoring system.

[0094] Feedback Data Collection: After feeding, the monitoring system collects four types of data in real time: remaining feed (the percentage of uneaten feed is counted through image recognition), changes in growth status (body length and weight growth rate), activity status response (changes in feeding activity and swimming frequency), and water quality-related data (changes in ammonia nitrogen and dissolved oxygen). At the same time, it also includes the mortality rate of fish fry and the incidence of diseases (such as the frequency of occurrence of diseases such as skin ulcers and gill rot).

[0095] Iteration cycle adjustment: dynamically set according to the growth stage of fish fry: 24 hours / cycle during the fry stage, 48-72 hours / cycle during the growth stage; when the parameter fluctuation exceeds 15% (such as the feeding activity fluctuation from 8 times / minute to 6.8 times / minute), shorten it to 12 hours / cycle;

[0096] Incremental learning optimization: The model is updated using an incremental learning algorithm, and new data is assigned a weight based on time: data within 7 days has a weight of 0.8, data from 8 to 30 days has a weight of 0.7, and data older than 30 days has a weight of 0.6; outliers (such as abnormal data where the mortality rate of fish fry suddenly increases by more than 5%) are removed using the 3σ principle.

[0097] Solution update mechanism: If the prediction accuracy of growth status and activity status parameters improves by ≥5% after iteration, the optimal feeding solution is updated; otherwise, a data anomaly report is generated to investigate equipment failures (such as sensor calibration deviations) or environmental problems (such as sudden changes in water quality) to ensure the effectiveness of closed-loop optimization.

[0098] 6. Monitoring system calibration and maintenance

[0099] To ensure the accuracy of the model input data and the optimization effect, the monitoring system is calibrated and maintained regularly.

[0100] Calibration cycle: once every 7-15 days; when the water quality is turbid (turbidity > 20 NTU) or the parameter fluctuation exceeds 20%, shorten to once every 7 days;

[0101] Calibration content:

[0102] Underwater HD camera: After cleaning the lens, calibrate it using a standard reference (such as a scale plate with known length) to ensure that the body length measurement error is ≤2%;

[0103] Biosensor array: Sensitivity is calibrated in a standard simulated environment (such as fish fry samples with known body length and weight, and standard water flow velocity) to ensure detection error ≤3%;

[0104] Data transmission guarantee: The signal strength of the data transmission module must be ≥-70dBm, and the wireless transmission success rate must be ≥99% to avoid data loss;

[0105] Aquaculture record management: Establish encrypted aquaculture records, record calibration data, model iteration logs, feeding plan adjustment records and fry growth reports, and store them in the cloud for ≥2 years to provide continuous historical data support for model training.

[0106] like Figure 2 As shown in the figure, a specific embodiment of the present invention provides a feed regulation system based on the growth status of fish fry, the system comprising:

[0107] This system serves as the hardware and software implementation platform for the above methods, and includes the following functional modules:

[0108] 1. Parameter acquisition module: It consists of an image acquisition device (underwater high-definition camera, infrared supplementary light module) and a biosensor array (body length and weight monitoring sensor, motion state sensor, feeding behavior sensor, and cluster distribution sensor), which is used to collect fish fry growth status monitoring parameters and activity status monitoring parameters in real time, and simultaneously collect water quality environmental parameters.

[0109] 2. Data Processing Module: Communicates with the parameter acquisition module, receives the collected parameter data, and uses data visualization algorithms to create growth trend charts, activity status distribution charts, and growth-activity correlation charts, presenting the correlation characteristics between fish fry growth and activity. It also supports parameter filtering and time period scaling.

[0110] 3. Feeding plan generation module: It communicates with the data processing module, analyzes the data information graph to extract key features, and formulates an initial feeding plan based on the basic nutritional requirements of the fish fry species;

[0111] 4. Feeding plan execution optimization module: It communicates with the feeding plan generation module, has a built-in feed regulation model constructed by gradient boosting decision tree algorithm, inputs historical data to train the model, establishes the mapping relationship between multiple parameters and feeding plan, outputs the best feeding plan, and controls the feeding equipment (such as automatic feeder) to execute the plan;

[0112] 5. Data Feedback Dynamic Optimization Module: It communicates with the parameter acquisition module and the feeding plan execution optimization module respectively, receives the feedback data after feeding, transmits it to the feed control model, triggers the model to iteratively optimize at regular intervals, updates the best feeding plan, and realizes closed-loop control.

[0113] Data transmission between modules is achieved via industrial Ethernet or wireless communication (such as LoRa, 5G), with a transmission delay of ≤10 seconds, ensuring timely system response. Detailed Implementation

[0115] Example: Feed Regulation for California bass fry

[0116] 1. Breeding conditions

[0117] Culture pond specifications: rectangular cement pond (20m long × 10m wide × 1.5m deep), culture area 200㎡, water depth 1.2m; fish fry species: California bass fry (initial body length 2cm, initial weight 0.5g), culture density 30 fish / ㎡, total 6000 fish; culture cycle: 30 days (1-15 days for fry, 16-30 days for growth).

[0118] 2. Monitoring System Deployment

[0119] Image acquisition device: Two underwater high-definition cameras (8 megapixels, 30fps, 10mm focal length) are deployed and installed in the upper, middle and lower layers of the middle of the aquaculture pond (0.3m, 0.6m and 0.9m from the bottom of the pond, respectively). Each camera has a coverage radius of 6 meters and no blind spots. The infrared supplementary lighting module has a wavelength of 900nm and is linked with the water transparency sensor.

[0120] Biosensor array: 4 groups are deployed (1 group per 50㎡), each group includes body length and weight monitoring sensors, motion status sensors, feeding behavior sensors, and cluster distribution sensors. Modular installation, data acquisition frequency: 0.5 hours / time during seedling stage, 1 hour / time during growth stage;

[0121] Water quality sensors: Deploy sensors for water temperature, dissolved oxygen, pH, ammonia nitrogen, nitrite, turbidity, and water flow velocity, and collect data synchronously with biosensors.

[0122] 3. Data Processing and Initial Solution Development

[0123] Monitoring parameter collection: During the seedling stage (1-15 days), parameters such as body length, weight, daily growth rate, swimming frequency, and feeding activity were collected at 0.5-hour intervals; on the 5th day, when the daily growth rate of body length was 12%, the collection interval was shortened to 0.5 hours.

[0124] Data chart creation:

[0125] Growth trend chart: The horizontal axis represents time (0.5-hour nodes), and the vertical axis displays body length, weight, and average daily growth rate. The fitted curve shows that the body length increased from 2cm to 3.5cm, and the peak average daily growth rate was 0.1cm / day (8:00-12:00).

[0126] Activity status distribution map: The peak range of feeding activity is 10:00-11:00 (12 times / minute) and 16:00-17:00 (10 times / minute).

[0127] Growth-activity correlation plot: Feeding activity level and weight gain rate are positively correlated by a coefficient of 0.88;

[0128] Initial feeding plan:

[0129] Nutritional requirements: Protein 52%, Fat 12%, Vitamin A:D:E:C = 3:1:2:5, Calcium 0.9%, Phosphorus 0.7%;

[0130] Feed particle size: The ratio of feed particle size to the average mouth diameter of fish fry (2mm) is 1:1.3, and high-protein feed with a particle size of 2.6mm is selected;

[0131] Single feeding amount: calculated at 5% of body weight (initial single feeding amount 0.025g / tail, twice a day, total feeding amount 300g / day);

[0132] Feeding times: 10:00-10:30, 16:00-16:30;

[0133] Feeding uniformity: density difference ≤8%.

[0134] 4. Feed regulation model training and optimal solution output

[0135] Environmental parameter control: water temperature 25℃, dissolved oxygen 6mg / L, pH 7.8, ammonia nitrogen 0.1mg / L, nitrite 0.05mg / L, turbidity 15NTU, water flow velocity 0.2m / s;

[0136] Model parameters: learning rate 0.05, decision tree depth (4 layers in the seedling stage and 7 layers in the growth stage), minimum number of samples per leaf node 10, regularization coefficient 0.005;

[0137] Historical data: Data from the first three batches of California bass fry rearing (90 days in total) were used, divided into a training set (63 days) and a test set (27 days) at a ratio of 7:3. After 5-fold cross-validation, the core parameter prediction accuracy was 94%.

[0138] Optimal feeding plan:

[0139] Seedling stage (1-15 days): High protein feed, single feeding amount is 5.2% of body weight, twice a day (10:05-10:25, 16:10-16:30).

[0140] Growth period (16-30 days): Balanced nutrition feed (40% protein), single feeding amount 3.8% of body weight, 3 times a day (8:30-8:50, 12:00-12:20, 16:30-16:50).

[0141] 5. Dynamically optimize execution

[0142] Feedback data collection: After feeding, collect the remaining food intake (≤3%), body length growth rate (average 0.12cm per day), and feeding activity level (11 times / minute).

[0143] Iterative optimization: Iterations were performed every 24 hours during the seedling stage and every 48 hours during the growth stage; on day 20, the stocking density was monitored to increase to 35 fish / m², triggering an 8% increase in feeding amount;

[0144] Farming results: After 30 days, the average body length of California bass fry was 5.8cm, the average weight was 3.2g, the body length difference was 8%, the disease incidence rate was 2%, the feed conversion ratio was 1.2 (0.3 lower than traditional feeding), and the survival rate was 96% (8% higher than traditional feeding).

[0145] The beneficial effects of this invention are as follows: By employing a monitoring method that integrates image acquisition and a biosensor array, this invention achieves non-contact, multi-dimensional, and blind-spot-free monitoring of growth and activity parameters, with a data acquisition error of ≤5%, solving the problem of inefficient traditional monitoring. Through multi-dimensional data chart analysis of the correlation between growth and activity, and combined with a gradient boosting decision tree algorithm to construct a control model, it achieves precise matching of feed type, feeding amount, and feeding time, with a core parameter prediction accuracy of ≥92%. It establishes a closed-loop mechanism of "feeding-monitoring-optimization," adjusting the feeding plan in real time according to changes in fry growth, activity response, and environmental fluctuations, adapting to different growth stages and aquaculture environments, and improving the adaptability of the plan. It reduces feed waste (residual feed percentage ≤5%), promotes uniform fry growth (body length difference ≤10%), reduces disease incidence (≤3%), saves labor costs, and improves survival rate and economic benefits. The monitoring system adopts a modular design, supports calibration, maintenance, and troubleshooting, with a data transmission success rate of ≥99%, and aquaculture records are stored in the cloud for ≥2 years, providing data support for long-term aquaculture.

[0146] The above descriptions are merely embodiments of the present invention. Commonly known technical solutions or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the scope of the present invention, and these should also be considered within the protection scope of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for regulating feed based on the growth status of fish fry, characterized in that, The method includes: S1: A monitoring system consisting of an image acquisition device and a biosensor array is used to monitor fish fry in aquaculture water in real time and obtain growth status monitoring parameters and activity status monitoring parameters. S2. Based on the growth status monitoring parameters and activity status monitoring parameters collected in S1, establish multiple data information graphs, including a growth trend graph, an activity status distribution graph, and a growth-activity correlation graph, to present the correlation characteristics between fish fry growth and activity. S3. Analyze the data information graph established in S2, extract the peak growth period, active feeding period and the correlation characteristics between the two, and formulate an initial feeding plan with clear feed type, single feeding amount, daily feeding frequency and time window in combination with the basic nutritional requirements of fish fry. S4. Based on the initial feeding scheme as a sample, multiple environmental parameters such as water temperature, dissolved oxygen, and pH are incorporated. A feed regulation model is constructed using a gradient boosting decision tree algorithm. Historical fish fry growth activity data and corresponding feeding effect data are input to train the feed regulation model. A mapping relationship between multiple parameters and the feeding scheme is established, and the optimal feeding scheme, including the best feed type, precise feeding amount, and optimal feeding time, is output. S5. Apply the optimal feeding scheme to actual aquaculture. The monitoring system collects data on the remaining amount of feed, growth and activity status of the fish fry after feeding in real time, and feeds it back to the feed control model. The feed control model is iteratively optimized at regular intervals to achieve dynamic optimization of the optimal feeding scheme.

2. The feed regulation method based on the growth status of fish fry according to claim 1, characterized in that: The image acquisition device includes an underwater high-definition camera and an infrared fill light module; The underwater high-definition camera adopts an IP68 waterproof and dustproof sealing design, with a pixel count of ≥8 million, a frame rate of 25-30fps, and a focal length of 8-12mm. Each unit can cover a radius of 5-8 meters in the aquaculture water body. The aquaculture pond is evenly distributed with a density of one unit per 100㎡ in three layers (upper, middle, and lower) to ensure no monitoring blind spots. The infrared supplementary light module has a wavelength range of 850-940nm. The irradiation intensity is dynamically adjusted in real time based on the feedback data from the water transparency sensor. When the transparency is ≤15cm, the irradiation intensity is increased by 50%, effectively covering more than 90% of the monitoring area. The biosensor array consists of body length and weight monitoring sensors, motion state sensors, feeding behavior sensors, and clustered distribution sensors. It adopts a modular plug-and-play design, and the data acquisition frequency is synchronized with the image acquisition device, and synchronization is performed once every 0.5-2 hours. The body length and weight sensors achieve non-contact detection through laser ranging and image recognition fusion technology. The motion state sensor captures the instantaneous swimming trajectory of the fish fry based on triaxial accelerometer technology. The static detection error of all sensors is ≤3%, and the dynamic detection error is ≤5%.

3. The feed regulation method based on the growth status of fish fry according to claim 1, characterized in that: The growth status monitoring parameters include body length, body weight, average daily growth rate, body width, condition factor, and scale development completeness: The body width is the lateral distance at the widest point of the fish fry's body; The formula for calculating the fullness is: ； in, For plumpness, The length of the fish fry, For the width of the fish fry, As a benchmark factor, and The product of these factors constitutes the phenotypic volume characteristic of the fish fry, which is then compared with the baseline factor. The ratio eliminates the influence of individual size differences on developmental assessment and realizes the quantification of "fatness / thinness"; The completeness of scale development is determined by statistical analysis of scale coverage and scale damage rate using image recognition. The activity status monitoring parameters include swimming frequency, feeding activity, cluster distribution density, stress response frequency, diurnal activity difference coefficient, and feeding response time; The data acquisition interval for the growth status monitoring parameters and the activity status monitoring parameters is 0.5-2 hours, with 0.5-1 hours during the seedling stage and 1-2 hours during the growth stage. When the daily growth rate of the body length exceeds 10% or the stress response frequency exceeds the limit three times consecutively, the data acquisition interval is shortened to 0.5 hours to ensure the timeliness of data acquisition during key growth stages.

4. The feed regulation method based on the growth status of fish fry according to claim 3, characterized in that: Using the body length, weight, and daily growth rate of the growth status monitoring parameters collected in step S1, and the swimming frequency, feeding activity, and cluster distribution density of the activity status monitoring parameters, as the core data sources, the coordinate axes and presentation logic of each chart are defined: The growth trend chart uses time as the horizontal axis and marks nodes at 1-2 hour collection intervals. The vertical axis simultaneously displays the values ​​of three types of growth parameters. Different colored broken lines are used to fit the data points, intuitively presenting the changing pattern of single parameters over time and the synergistic growth / fluctuation characteristics of multiple parameters. The activity status distribution chart uses the activity parameter type as the horizontal axis and the frequency proportion of the corresponding parameter as the vertical axis. It adopts the form of bar chart + error bar and marks the mean, standard deviation and peak range of each parameter. The growth-activity correlation plot is presented in the form of a dual-axis scatter plot, with the horizontal axis representing feeding activity, the left vertical axis representing the average daily growth rate, and the right vertical axis representing the weight gain rate. The positive / negative correlation between the two is quantified by the degree of clustering of data points and the slope of the fitted curve. The growth trend chart, activity status distribution chart, and growth-activity correlation chart all support parameter filtering and time period scaling.

5. The feed regulation method based on the growth status of fish fry according to claim 1, characterized in that: The basic nutritional requirements for the fish fry species include: clearly quantified indicators: protein 35%-55%, fat 8%-15%, vitamin A: vitamin D: vitamin E: vitamin C = 3:1:2:5, and minerals calcium ≥0.8% and phosphorus ≥0.6%; The initial feeding plan is as follows: the ratio of feed pellet size to the average mouth diameter of the fish fry is 1:1.2-1:1.5; feeding uniformity is randomly detected by the monitoring system to ensure density difference is ≤10%; and the duration of each feeding session is 15-30 minutes. All parameters are calibrated based on the characteristics of the fish fry species, growth stage, and data from at least three previous feeding and rearing batches.

6. The feed regulation method based on the growth status of fish fry according to claim 3, characterized in that: When constructing the feed regulation model, the environmental parameters also include controlling ammonia nitrogen concentration ≤0.2mg / L, nitrite concentration ≤0.1mg / L, water turbidity ≤20NTU, and water flow velocity in the middle layer of the water body to be 0.1-0.3m / s; The growth status monitoring parameters and activity status monitoring parameters are collected synchronously using a dedicated water quality sensor. The collection frequency is consistent with that of the growth status monitoring parameters. The data transmission delay is ≤10 seconds. After collection, the data is first normalized to map it to the [0,1] interval to remove extreme value interference. The gradient boosting decision tree algorithm dynamically adapts its parameters as follows: the initial learning rate is 0.05, which is reduced to 0.03 when the training set fitting error is >5%, and increased to 0.08 when the fitting error is <2%. The depth of the decision tree is adjusted according to the growth stage of the fish fry, with 3-5 layers in the fry stage and 6-8 layers in the growth stage. The minimum number of samples in the leaf nodes is ≥10, and the regularization coefficient is 0.005 to suppress overfitting. The training steps of the feed regulation model are as follows: historical fish fry feeding data are divided into training set and test set in a 7:3 ratio. The training set is cross-validated with 5 folds. The data in each fold is randomly divided and repeated 3 times. The average value is taken to optimize the parameters. The feed regulation model has a prediction accuracy of ≥92% for the core parameters of the optimal feeding scheme. The prediction deviation of the average daily growth rate and the feeding activity is ≤8%, and the deviation of non-core parameters is ≤12%, providing stable algorithm support for the dynamic iterative optimization of S5.

7. The feed regulation method based on the growth status of fish fry according to claim 1, characterized in that: The optimal feeding plan consists of the core parameters output by the feed regulation model, including the optimal feed type that matches the nutritional needs of the fry, the precise feeding amount calculated based on the weight percentage, the optimal feeding time based on the active feeding period, and the feeding method that matches the appropriate density. The optimal feeding scheme is set with an automatic switching mechanism: when the stocking density exceeds 50 fish / m² or the daily growth rate of body length is greater than 8%, the feeding amount is dynamically adjusted; the pause condition is that the feeding activity is less than 3 times / minute for 5 consecutive minutes or the remaining feed is more than 20%. In the emergency scenario, the feeding is quickly adjusted according to the growth status monitoring parameters and the activity status monitoring parameters, and the monitoring system of S5 is linked to collect data on the remaining feed amount, growth and activity status changes of the fish fry after feeding in real time.

8. The feed regulation method based on the growth status of fish fry according to claim 1, characterized in that, The steps of applying the optimal feeding scheme to actual aquaculture, using the monitoring system to collect real-time data on the remaining feed intake, growth, and activity status of fish fry after feeding, and feeding this data back to the feed control model, where the feed control model is periodically iteratively optimized to achieve dynamic optimization of the optimal feeding scheme, include: After feeding according to the optimal feeding plan, four types of feedback data are collected through the monitoring system: remaining feed amount, changes in growth status, activity status response and water quality related data, while also including fry mortality rate and disease incidence rate. The iteration cycle of the feed regulation model is dynamically adjusted according to data fluctuations: 24 hours / time during the seedling stage, 48-72 hours / time during the growth stage, and shortened to 12 hours when the parameter fluctuation exceeds 15%. The iterative optimization process employs an incremental learning algorithm, with new data updated according to time weights: 0.8 for data within 7 days, 0.7 for data between 8 and 30 days, and 0.6 for data over 30 days. Outliers are eliminated using the 3σ principle. If the prediction accuracy of the growth status monitoring parameters and the activity status monitoring parameters improves by ≥5% after iteration, the optimal feeding scheme is updated; otherwise, a data anomaly report is generated to investigate equipment or environmental issues, thus forming a closed loop of "feeding-monitoring-optimization".

9. The feed regulation method based on the growth status of fish fry according to claim 2, characterized in that, Following step S5, the following is also included: To ensure the accuracy of the model input data and the optimization effect of S5, the monitoring system is calibrated and maintained every 7-15 days. When the water quality is turbid or the parameter fluctuation exceeds 20%, the calibration and maintenance time is shortened to 7 days. The calibration and maintenance includes cleaning the lens of the underwater high-definition camera and calibrating it with a standard reference to ensure that the body length measurement error is ≤2%. The biosensor array was calibrated for sensitivity in a standard simulated environment, with a detection error ≤3%. The signal strength of the data transmission module that transmits the growth status monitoring parameters and activity status monitoring parameters collected by the monitoring system to the computer must be ≥-70dBm, and the wireless transmission success rate must be ≥99%. Establish encrypted aquaculture records, record calibration data, model iteration logs, feeding scheme adjustment records, and fry growth reports, and store them in the cloud for ≥2 years to provide historical data support for the training of the feed regulation model and achieve continuous optimization of the optimal feeding scheme.

10. A feed regulation system based on the growth status of fish fry, characterized in that, The system includes a server, multiple sensor nodes, and a communication module, and includes the following functional modules: Parameter acquisition module: The monitoring system is composed of an image acquisition device and a biosensor array to monitor fish fry in aquaculture water in real time and acquire growth status monitoring parameters and activity status monitoring parameters; Data processing module: Based on the collected growth status monitoring parameters and activity status monitoring parameters, it establishes multiple data information graphs, including a growth trend graph, an activity status distribution graph, and a growth-activity correlation graph, to present the correlation characteristics between fish fry growth and activity. Feeding plan generation module: Analyze the established data information graph, extract the peak growth period, active feeding period and the correlation characteristics between the two, and formulate an initial feeding plan with clear feed type, single feeding amount, daily feeding frequency and time window in combination with the basic nutritional requirements of fish fry. Feeding scheme execution optimization module: Based on the initial feeding scheme as the base sample, multiple environmental parameters such as water temperature, dissolved oxygen, and pH value are incorporated. A feed regulation model is constructed using a gradient boosting decision tree algorithm. The feed regulation model is trained by inputting historical fish fry growth activity data and corresponding feeding effect data, establishing a mapping relationship between multiple parameters and the feeding scheme, and outputting the optimal feeding scheme including the best feed type, precise feeding amount, and optimal feeding time. Data feedback dynamic optimization module: The optimal feeding scheme is applied to actual aquaculture. The monitoring system collects data on the remaining amount of feed, growth and activity status of fish fry after feeding in real time, and feeds it back to the feed control model. The feed control model is iteratively optimized at regular intervals to achieve dynamic optimization of the optimal feeding scheme. The server is used to coordinate the data processing and storage of each module, and the sensor node includes an image acquisition device and a biosensor array.

Images