Aquaculture intelligent feeding management system based on Internet of Things

Abstract

The invention discloses an aquaculture intelligent feeding management system based on the Internet of Things, and relates to the technical field of aquaculture, a camera and a dissolved oxygen detector are utilized to construct an Internet of Things monitoring module, fish school behaviors and water quality environment data are collected in real time, and the movement track, group density and feeding state of a fish school are extracted; secondly, light-weight algorithms such as TinyYOLO and the like are adopted, rapid analysis of feeding intensity and feeding requirements is achieved in a local embedded computing module, a feeding model is generated, and dependence on high-performance hardware and cloud computing is reduced; meanwhile, the feeding control logic device dynamically calculates the optimal feeding amount and the optimal feeding opportunity based on the feeding state and the environmental characteristics; by monitoring the change of the ingestion intensity, the system catches the optimal ingestion window of the fish school, and maximization of the feed utilization rate is achieved; meanwhile, in combination with feedback analysis and a closed-loop optimization mechanism, model parameters are dynamically adjusted according to the feeding effect of the last round, and the environmental adaptability of the feeding plan is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aquaculture, and in particular to an intelligent feeding management system for aquaculture based on the Internet of Things. Background Art

[0002] At present, mainstream large-scale aquaculture feeding equipment mostly combines high-precision cameras, algorithm models and multiple sensors to work together to monitor fish behavior, water quality parameters and feeding effects. This type of equipment is expensive and complex to operate and maintain. For small and medium-sized farms or individual farmers with limited funds, the threshold for intelligent management is high.

[0003] In contrast, traditional aquaculture feeding management methods, such as scheduled feeding and feeding under manual observation, although requiring less equipment investment, are prone to problems such as under- or over-feeding, which not only causes feed waste and water pollution, but may also lead to uneven growth of fish. To solve this problem, some low-cost feeding solutions have attempted to use a single sensor for monitoring and automated feeding, but they perform poorly in dealing with complex farming conditions, especially in scenarios with large fluctuations in water quality or diverse fish behaviors. The results are unsatisfactory.

[0004] In addition, existing intelligent feeding solutions generally have the problem of high hardware dependence. Many systems use industrial-grade computing units and high-precision sensors, which not only increase the difficulty of system installation, debugging and maintenance, but also rely heavily on professional technology, further increasing operation and maintenance costs; even if some systems try to transfer computing tasks to the cloud, they face the problems of high data transmission costs and unstable network coverage in rural areas, making it particularly difficult to ensure feeding accuracy and efficiency while reducing input costs. Therefore, there is an urgent need for an intelligent feeding management system for aquaculture based on the Internet of Things to solve such problems. Summary of the invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides an intelligent feeding management system for aquaculture based on the Internet of Things to solve the problems in traditional aquaculture feeding schemes, namely, timed feeding and manual observation. Although the equipment investment is small, it is easy to cause insufficient or excessive feeding, resulting in feed waste, water pollution and uneven growth of fish schools. The low-cost single-sensor solution is not effective under complex conditions and is difficult to adapt to water quality fluctuations and diversified fish behavior scenarios, which limits its application effect.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] The embodiment of the present invention provides an intelligent feeding management system for aquaculture based on the Internet of Things, which includes:

[0009] IoT monitoring module, used to collect aquaculture data and extract fish characteristics data, including cameras and dissolved oxygen detectors, with built-in edge computing units;

[0010] The local embedded computing module is used to receive the fish characteristic data transmitted by the IoT monitoring module, use lightweight algorithms to analyze the feeding status of the fish, and generate a feeding intensity and feeding demand model; it uses an embedded processor and is equipped with a lightweight target detection algorithm;

[0011] A feeding control logic unit, including a control logic circuit and a calculation unit, is used to perform logic operations and feeding execution;

[0012] In the feeding control logic, the feeding amount and feeding timing are calculated based on the feeding status and demand model output by the local embedded computing module and combined with the actual characteristic data;

[0013] The feeder is used to receive instructions from the feeding control logic device, distribute feed according to the set feeding amount and time, and deliver it to the breeding area;

[0014] The feeder includes a feed storage bin and a feeding execution device. The feeding execution device uses a rotating feeding tray and a track-type feeding device.

[0015] The feeding data storage unit uses an embedded memory to store the feeding data of each feeding task, including feeding amount, time, feeding status analysis results and breeding data.

[0016] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the camera collects image data of the feeding behavior of the fish school in real time to capture the movement trajectory, group density and feeding status of the fish school;

[0017] The dissolved oxygen detector monitors the dissolved oxygen level and water temperature in the fish pond where the fish are located.

[0018] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the management method of the aquaculture intelligent feeding management system includes:

[0019] Step S1, using a camera and a dissolved oxygen detector to collect aquaculture data, and extracting fish school characteristic data based on the aquaculture data;

[0020] Step S2, the fish school characteristic data is transmitted to the local embedded computing module, and the lightweight target detection TinyYOLO algorithm is used to analyze the feeding status of the fish school to generate a feeding intensity and feeding requirement model;

[0021] Step S3, taking the analysis result of the feeding state as input, combining the actual feeding demand and characteristic data, and using the feeding control logic to calculate the feeding amount and timing combination;

[0022] Step S4, based on the calculation result of step S3, setting the automatic feeding instruction of the feeder, the feeding instruction matches the calculated feeding amount and feeding timing combination;

[0023] Step S5, after completing the feeding task, the effect of the previous round of feeding is fed back and analyzed in combination with the changes in the feeding status and characteristic data of the fish school, and the analysis results are used to optimize the next round of feeding plan.

[0024] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the farming data includes the movement trajectory, group density and feeding status of the fish school, and the dissolved oxygen and water temperature of the water quality.

[0025] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the step of extracting fish school characteristic data based on aquaculture data is as follows:

[0026] Use the camera to collect time series images and generate a frame sequence {I1, I2, ..., I n}, where n is the total number of frames, I k Indicates time t k The image frames collected at the same time, k = 1, 2, ..., n, for each frame of image I k Grayscale processing is performed to convert color pixels into grayscale pixels. The grayscale formula is:

[0027] G(x, y) = 0.299R(x, y) + 0.587G(x, y) + 0.114B(x, y), where n is the total number of frames in the time series, I k is time t k The image frame collected at , G(x, y) is the gray value of the pixel (x, y), R(x, y), G(x, y), B(x, y) are the values ​​of the pixel in the red, green, and blue channels, 0.299, 0.587, 0.114 are the weighting coefficients of the red, green, and blue channels to the gray value respectively;

[0028] The optical flow method is used to calculate the motion vector u(x, y, t k ), the calculation formula is:

[0029]

[0030] Solve for the velocity vector: u(x, y, t k )=(u, v), where u(x, y, t k) is the pixel point (x, y) at time t k The motion vector, u and v are the velocity components of the pixel along the x and y directions,

[0031] is the spatial gradient of the image,

[0032] is the temporal gradient of the image;

[0033] Use background subtraction to extract the fish school area, count the number of pixels in the target area and calculate the school density. The extraction formula is:

[0034]

[0035] Among them, ρ is the fish density, B(x, y) is a binary mask, the fish pixels are 1, the background pixels are 0, R is the fish target area, and A R is the area of ​​the target region;

[0036] The dissolved oxygen detector records water quality parameters at time intervals Δt to form a sequence of dissolved oxygen and water temperature, expressed as: Among them, W(t k ) is time t k The water quality characteristic vector at time , is time t k The dissolved oxygen concentration at water (t k ) is time t k The water temperature at that time, Δt is the sampling time interval of the dissolved oxygen detector;

[0037] Merge feature data into time t k The eigenvector F(t k ),

[0038]

[0039] Among them, F(t k ) is time t k The fish school characteristic vector, u(x, y, t k ) is the fish school motion vector, ρ(t k ) is the population density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature.

[0040] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the steps of using the lightweight target detection TinyYOLO algorithm to analyze the feeding state of the fish school and generate the feeding intensity and feeding demand model are as follows:

[0041] The fish school image feature F(t k ) is used as input for TinyYOLO to detect objects.

[0042] Defining inputs includes:

[0043] I ROI (t k ), is the time t k The image of the region of interest at that time, the cropped image range contains the feeding behavior of the fish school,

[0044] H×W×c is the image size, which is height H, width W, and number of channels C=1;

[0045] Using the TinyYOLO network structure, image I ROI (t k ) to extract features and generate a bounding box prediction set: in, is time t k is the bounding box set of the fish school target, b i =(x i ,y i , w i ,h i , p i ) is the feature of a single bounding box, where x i ,y i is the center coordinate of the bounding box, w i ,h i is the bounding box width and height, p i is the confidence level, m is the total number of bounding boxes detected;

[0046] Based on the bounding box set Calculate the feeding intensity using the formula:

[0047]

[0048] Among them, S(t k ) is time t k The feeding intensity at that time, A ROI is the area of ​​the region of interest, p i is the bounding box b i The confidence level, A(b i )=w i ·h i is the bounding box b i The feeding activity of the fish school in the ROI is calculated by weighting the confidence and area of ​​all bounding boxes;

[0049] Using feeding intensity S(t k) and environmental parameters to generate a feeding demand model, the model formula is:

[0050]

[0051] Among them, N(t k ) is time t k The feeding requirement at that time, S(t k ) is time t k The intensity of feeding, is time t k The dissolved oxygen concentration at water (t k ) is time t k α, β, γ are the weight parameters of feeding intensity, dissolved oxygen and water temperature;

[0052] The feeding intensity model S(t k ), describing the feeding activity of the fish at a specific moment, and the feeding demand model N(t k ), provide quantitative feeding amount requirements.

[0053] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the step of taking the analysis result of the feeding state as input, combining the actual feeding demand and characteristic data, and using the feeding control logic to calculate the feeding amount and timing combination is as follows:

[0054] Acquiring data from step S2 includes:

[0055] Feeding intensity S(t k ), feeding requirement N(t k ) and the environmental characteristic vector E(t k ),

[0056] in, is the dissolved oxygen concentration, θ water (t k ) is the water temperature;

[0057] According to the feeding intensity S(t k ), environmental characteristics E(t k ), dynamic calculation time t k Feeding amount M(t k ), the calculation formula is:

[0058]

[0059] Among them, M(t k ) is time t k Feeding amount, N(t k ) is the feeding requirement, is the dissolved oxygen concentration, θ water (t k ) is water temperature, λ1, λ2, λ3 are weight coefficients of feeding demand, dissolved oxygen and water temperature;

[0060] The feeding time is defined as the time window that meets the feeding needs of the fish. The feeding intensity change trend ΔS(t k ) to determine the optimal time, the calculation formula is:

[0061] ΔS(t k )=S(t k )-S(t k-1 ),

[0062] Trigger conditions for feeding time:

[0063] If ΔS(t k )>θ S and Then T(t k ) = t k ,

[0064] Otherwise, T(t k ) is the delay, where T(t k ) is time t k The feeding trigger time, ΔS(t k ) is the change in feeding intensity, θ S is the feeding intensity threshold, is the dissolved oxygen concentration threshold;

[0065] The final feeding plan includes the combination of feeding amount and time, and the formula is:

[0066] P(t k )={M(t k ), T(t k )},

[0067] Among them, P(t k ) is time t k The feeding plan, M(t k ) is the feeding amount, T(t k ) is the feeding time.

[0068] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the step of setting the automatic feeding instruction of the feeder based on the calculation result of step S3 is as follows:

[0069] The feeding plan P(t k ), including: feeding amount M(t k ) and feeding time T(t k );

[0070] Generate feeding instructions, feeding instructions C(t k ) including: feeding start time and feeding equipment operating parameters,

[0071] The formula for generating feeding instructions is:

[0072] C(t k )={t s , v f , d s}, where C(t k ) is time t k Feeding instruction, t s =T(t k ) is the start time of the feeder,

[0073] is the feeding speed of the feeder, where

[0074] M(t k ) is the feeding amount, τ d The duration of feeding is preset according to the specifications of the feeder.

[0075] is the feeding coverage density, where r is the rotation range of the feeder, A t is the target area for feeding;

[0076] The feeding command is sent to the feeder execution device, including:

[0077] Feeding controller, used to start or stop feed feeding and control feeding speed v f and duration τ d ,

[0078] The rotating throwing device is used to t and feeding coverage density d s , dynamically adjust the rotation speed;

[0079] The feeder performs operations according to the received instructions and monitors the feeding process through sensors. The execution process is as follows:

[0080]

[0081] Among them, M act (t k ) is the actual feeding amount, t s is the feeding start time, τ d is the feeding duration, v f (t) is the actual material feeding speed;

[0082] The feeder will feed the actual amount Mact (t k ), the feeding coverage and time are fed back to the feeding data storage unit, and the feedback record is: R(t k )={M act (t k ), d s , t s}, where R(t k ) is time t k Feeding records, M act (t k ) is the actual feeding amount, d s is the feeding coverage density, t s The actual startup time.

[0083] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the step of combining the changes in the feeding status and characteristic data of the fish school to conduct feedback analysis on the effect of the previous round of feeding, and using the analysis results to optimize the next round of feeding plan is as follows:

[0084] Obtain the actual feeding record R(t k ) and feeding status data S(t k ), and combined with the fish school characteristic data F(t k ) and environmental data E(t k ), forming the input for feedback analysis:

[0085]

[0086] Among them, A(t k ) is time t k Feedback analysis input data, M act (t k ) is the actual feeding amount, S(t k ) is the feeding intensity, ρ(t k ) is the fish density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature;

[0087] Compare the actual feeding intensity S(t k ) and expected feeding intensity S exp (t k ) to evaluate the feeding effect and calculate the feeding deviation. The calculation formula is:

[0088] ΔS(t k )=S(t k )-S exp (t k ),

[0089] Among them, ΔS(t k ) is time t k Deviation of feeding intensity, S exp (t k ) is time t k The expected feeding intensity,

[0090] The actual feeding amount deviation calculation formula is:

[0091] ΔM(t k )=M act (t k )-M(t k ),

[0092] Among them, ΔM(t k ) is time t k Feeding amount deviation, M(t k ) is the planned feeding amount.

[0093] As a preferred solution of the aquaculture intelligent feeding management system based on the Internet of Things described in the present invention, the step of performing feedback analysis on the effect of the previous round of feeding and using the analysis results to optimize the next round of feeding plan also includes:

[0094] Comprehensively consider the impact of environmental parameters on feeding behavior and define the environmental correction weight ω(t k ), the correction formula is:

[0095]

[0096] Among them, ω(t k ) is the environmental correction weight, α, β are the weight coefficients of dissolved oxygen and water temperature,

[0097] The corrected feeding bias is:

[0098] ΔS adj (t k )=ΔS(t k )+ω(t k );

[0099] Update the feeding model parameters based on the feedback results and optimize the next round of feeding plan. The optimization formula is:

[0100] M opt (t k+1 )=M(t k )+λ1ΔS adj (t k )-λ2ΔM(t k ),

[0101] Among them, M opt (t k+1 ) is time tk+1 The optimal feeding amount, λ1, λ2 are the adjustment coefficients of feeding deviation and feeding deviation,

[0102] Generate the next round of feeding plan:

[0103] P(t k+1 )={M opt (t k+1 ), T opt (t k+1 )}, where feeding time T opt (t k+1 ) is recalculated by analyzing the trend of feeding intensity changes, and the calculation formula is:

[0104]

[0105] The beneficial effects of the present invention are:

[0106] The present invention uses a camera and a dissolved oxygen detector to construct an Internet of Things monitoring module to collect fish behavior and water quality environment data in real time, and extract the movement trajectory, group density and feeding status of the fish; secondly, lightweight algorithms such as TinyYOLO are used to realize rapid analysis of feeding intensity and feeding needs in a local embedded computing module, generate a feeding model, reduce dependence on high-performance hardware and cloud computing, reduce system costs, and improve data processing efficiency.

[0107] In the present invention, based on the feeding status and environmental characteristics, the feeding control logic dynamically calculates the optimal feeding amount and feeding time, so that the feeding decision can flexibly adapt to the needs of the fish school and environmental changes, and avoid the rigidity of regular and quantitative feeding; by monitoring the changes in feeding intensity, the system captures the optimal feeding window of the fish school and maximizes the feed utilization rate; at the same time, combined with feedback analysis and closed-loop optimization mechanism, the system dynamically adjusts the model parameters according to the effect of the previous round of feeding, further improving the accuracy and environmental adaptability of the feeding plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0109] Figure 1 It is a structural schematic diagram of the aquaculture intelligent feeding management system of the present invention. DETAILED DESCRIPTION

[0110] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0111] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0112] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0113] Example 1, reference Figure 1 This embodiment provides an aquaculture intelligent feeding management system based on the Internet of Things, including:

[0114] IoT monitoring module, used to collect aquaculture data and extract fish characteristics data, including cameras and dissolved oxygen detectors, with built-in edge computing units;

[0115] The camera collects real-time image data of fish feeding behavior, capturing the movement trajectory, group density and feeding status of the fish;

[0116] Dissolved oxygen detectors monitor dissolved oxygen levels and water temperature in the fish ponds where the fish are kept;

[0117] The local embedded computing module is used to receive the fish characteristic data transmitted by the IoT monitoring module, use lightweight algorithms to analyze the feeding status of the fish, and generate a feeding intensity and feeding demand model; it uses an embedded processor and is equipped with a lightweight target detection algorithm;

[0118] A feeding control logic unit, including a control logic circuit and a calculation unit, is used to perform logic operations and feeding execution;

[0119] In the feeding control logic, the feeding amount and feeding timing are calculated based on the feeding status and demand model output by the local embedded computing module and combined with the actual characteristic data;

[0120] The feeder is used to receive instructions from the feeding control logic device, distribute feed according to the set feeding amount and time, and deliver it to the breeding area;

[0121] The feeder includes a feed storage bin and a feeding execution device. The feeding execution device uses a rotating feeding tray and a track-type feeding device.

[0122] The feeding data storage unit adopts an embedded memory to store the feeding data of each feeding task, including feeding amount, time, feeding status analysis results and breeding data;

[0123] This embodiment also provides a management method for an aquaculture intelligent feeding management system based on the Internet of Things, including:

[0124] Step S1, using a camera and a dissolved oxygen detector to collect aquaculture data, and extracting fish school characteristic data based on the aquaculture data;

[0125] Farming data include fish movement trajectory, population density and feeding status, as well as water quality dissolved oxygen and water temperature;

[0126] The steps for extracting fish characteristics data based on aquaculture data are as follows:

[0127] Use the camera to collect time series images and generate a frame sequence {I1, I2, ..., I n}, where n is the total number of frames, I k Indicates time t k The image frames collected at the same time, k = 1, 2, ..., n, for each frame of image I k Grayscale processing is performed to convert color pixels into grayscale pixels. The grayscale formula is:

[0128] G(x, y) = 0.299R(x, y) + 0.587G(x, y) + 0.114B(x, y), where n is the total number of frames in the time series, I k is time t k The image frame collected at , G(x, y) is the gray value of the pixel (x, y), R(x, y), G(x, y), B(x, y) are the values ​​of the pixel in the red, green, and blue channels, 0.299, 0.587, 0.114 are the weighting coefficients of the red, green, and blue channels to the gray value respectively;

[0129] The optical flow method is used to calculate the motion vector u(x, y, t k ), the calculation formula is:

[0130]

[0131] Solve for the velocity vector: u(x, y, t k )=(u, v), where u(x, y, t k ) is the pixel point (x, y) at time t k The motion vector, u and v are the velocity components of the pixel along the x and y directions,

[0132] is the spatial gradient of the image,

[0133] is the temporal gradient of the image;

[0134] Use background subtraction to extract the fish school area, count the number of pixels in the target area and calculate the school density. The extraction formula is:

[0135]

[0136] Among them, ρ is the fish density, B(x, y) is a binary mask, the fish pixels are 1, the background pixels are 0, R is the fish target area, and A R is the area of ​​the target region;

[0137] The dissolved oxygen detector records water quality parameters at time intervals Δt to form a sequence of dissolved oxygen and water temperature, expressed as: Among them, W(t k ) is time t k The water quality characteristic vector at time , is time t k The dissolved oxygen concentration at water (t k ) is time t k The water temperature at that time, Δt is the sampling time interval of the dissolved oxygen detector;

[0138] Merge feature data into time t k The eigenvector F(t k ),

[0139]

[0140] Among them, F(t k ) is time t k The fish school characteristic vector, u(x, y, t k ) is the fish school motion vector, ρ(t k ) is the population density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature;

[0141] Specifically, in this step, all features are seamlessly integrated with water quality data through images.

[0142] Step S2, the fish school characteristic data is transmitted to the local embedded computing module, and the lightweight target detection TinyYOLO algorithm is used to analyze the feeding status of the fish school to generate a feeding intensity and feeding requirement model;

[0143] The lightweight target detection TinyYOLO algorithm is used to analyze the feeding status of fish schools. The steps to generate the feeding intensity and feeding requirement model are as follows:

[0144] The fish school image feature F(t k ) is used as input for TinyYOLO to detect objects.

[0145] Defining inputs includes:

[0146] I ROI (t k ), is the time t k The image of the region of interest at that time, the cropped image range contains the feeding behavior of the fish school,

[0147] H×W×c is the image size, which is height H, width W, and number of channels c=1;

[0148] Using the TinyYOLO network structure, image I ROI (t k ) to extract features and generate a bounding box prediction set: in, is time t k is the bounding box set of the fish school target, b i =(x i ,y i , w i ,h i , p i ) is the feature of a single bounding box, where x i ,y i is the center coordinate of the bounding box, w i ,h i is the bounding box width and height, p i is the confidence level, m is the total number of bounding boxes detected;

[0149] Based on the bounding box set Calculate the feeding intensity using the formula:

[0150]

[0151] Among them, S(t k ) is time t k The feeding intensity at that time, A ROI is the area of ​​the region of interest, p i is the bounding box b i The confidence level, A(b i )=w i ·h i is the bounding box b i The feeding activity of the fish school in the ROI is calculated by weighting the confidence and area of ​​all bounding boxes;

[0152] Using feeding intensity S(t k) and environmental parameters to generate a feeding demand model, the model formula is:

[0153]

[0154] Among them, N(t k ) is time t k The feeding requirement at that time, S(t k ) is time t k The intensity of feeding, is time t k The dissolved oxygen concentration at water (t k ) is time t k α, β, γ are the weight parameters of feeding intensity, dissolved oxygen and water temperature;

[0155] The feeding intensity model S(t k ), describing the feeding activity of the fish at a specific moment, and the feeding demand model N(t k ), provide quantitative feeding amount requirements;

[0156] Specifically, the TinyYOLO network is used to reduce the number of parameters and calculations, which is suitable for embedded local computing environments. By combining feeding intensity with environmental characteristics, feeding requirements are dynamically adjusted to optimize resource utilization, and through bounding box confidence and area calculations, rapid quantification of fish feeding activity is achieved.

[0157] Step S3, taking the analysis result of the feeding state as input, combining the actual feeding demand and characteristic data, and using the feeding control logic to calculate the feeding amount and timing combination;

[0158] The analysis results of feeding status are used as input, combined with actual feeding needs and characteristic data, and the steps of calculating the feeding amount and timing combination using feeding control logic are as follows:

[0159] Acquiring data from step S2 includes:

[0160] Feeding intensity S(t k ), feeding requirement N(t k ) and the environmental characteristic vector E(t k ),

[0161] in, is the dissolved oxygen concentration, θ water (t k ) is the water temperature:

[0162] According to the feeding intensity S(t k ), environmental characteristics E(t k ), dynamic calculation time t kFeeding amount M(t k ), the calculation formula is:

[0163]

[0164] Among them, M(t k ) is time t k Feeding amount, N(t k ) is the feeding requirement, is the dissolved oxygen concentration, θ water (t k ) is water temperature, λ1, λ2, λ3 are weight coefficients of feeding demand, dissolved oxygen and water temperature;

[0165] The feeding time is defined as the time window that meets the feeding needs of the fish. The feeding intensity change trend ΔS(t k ) to determine the optimal time, the calculation formula is:

[0166] ΔS(t k )=S(t k )-S(t k-1 ),

[0167] Trigger conditions for feeding time:

[0168] If ΔS(t k )>θ S and Then T(t k ) = t k ,

[0169] Otherwise, T(t k ) is the delay, where T(t k ) is time t k The feeding trigger time, ΔS(t k ) is the change in feeding intensity, θ S is the feeding intensity threshold, is the dissolved oxygen concentration threshold;

[0170] The final feeding plan includes the combination of feeding amount and time, and the formula is:

[0171] P(t k )={M(t k ), T(t k )},

[0172] Among them, P(t k ) is time t k The feeding plan, M(t k ) is the feeding amount, T(t k ) is the feeding time;

[0173] Specifically, by combining feeding status with environmental characteristics, precise feeding can be achieved to avoid waste and water pollution, and based on the changing trend of feeding intensity, the best feeding window of fish schools can be captured to optimize feed utilization.

[0174] Step S4, based on the calculation result of step S3, setting the automatic feeding instruction of the feeder, the feeding instruction matches the calculated feeding amount and feeding timing combination;

[0175] Based on the calculation result of step S3, the step of setting the automatic feeding instruction of the feeder is as follows:

[0176] The feeding plan P(t k ), including: feeding amount M(t k ) and feeding time T(t k );

[0177] Generate feeding instructions, feeding instructions C(t k ) including: feeding start time and feeding equipment operating parameters,

[0178] The formula for generating feeding instructions is:

[0179] C(t k )={t s , v f , d s}, where C(t k ) is time t k Feeding instruction, t s =T(t k ) is the start time of the feeder,

[0180] is the feeding speed of the feeder, where

[0181] M(t k ) is the feeding amount, τ d The duration of feeding is preset according to the specifications of the feeder.

[0182] is the feeding coverage density, where r is the rotation range of the feeder, A t is the target area for feeding;

[0183] The feeding command is sent to the feeder execution device, including:

[0184] Feeding controller, used to start or stop feed feeding and control feeding speed v f and duration τ d ,

[0185] The rotating throwing device is used to tand feeding coverage density d s , dynamically adjust the rotation speed;

[0186] The feeder performs operations according to the received instructions and monitors the feeding process through sensors. The execution process is as follows:

[0187]

[0188] Among them, M act (t k ) is the actual feeding amount, t s is the feeding start time, τ d is the feeding duration, v f (t) is the actual material feeding speed;

[0189] The feeder will feed the actual amount M act (t k ), the feeding coverage and time are fed back to the feeding data storage unit, and the feedback record is: R(t k )={M act (t k ), d s , t s}, where R(t k ) is time t k Feeding records, M act (t k ) is the actual feeding amount, d s is the feeding coverage density, t s is the actual start time;

[0190] Specifically, based on the feeding amount and regional density calculation, dynamic adjustment of the feeder is achieved. By monitoring the actual feeding amount, the feeding behavior is consistent with the plan to reduce waste.

[0191] Step S5, after completing the feeding task, combining the changes in the feeding status and characteristic data of the fish school, feedback analysis is performed on the effect of the previous round of feeding, and the analysis results are used to optimize the next round of feeding plan;

[0192] Combined with the changes in the feeding status and characteristic data of the fish school, feedback analysis is conducted on the effect of the previous round of feeding. The analysis results are used to optimize the next round of feeding plan.

[0193] Obtain the actual feeding record R(t k ) and feeding status data S(t k ), and combined with the fish school characteristic data F(t k ) and environmental data E(t k ), forming the input for feedback analysis:

[0194]

[0195] Among them, A(t k ) is time t k Feedback analysis input data, M act (t k ) is the actual feeding amount, S(t k ) is the feeding intensity, ρ(t k ) is the fish density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature;

[0196] Compare the actual feeding intensity S(t k ) and expected feeding intensity S exp (t k ) to evaluate the feeding effect and calculate the feeding deviation. The calculation formula is:

[0197] ΔS(t k )=S(t k )-S exp (t k ),

[0198] Among them, ΔS(t k ) is time t k Deviation of feeding intensity, S exp (t k ) is time t k The expected feeding intensity,

[0199] The actual feeding amount deviation calculation formula is:

[0200] ΔM(t k )=M act (t k )-M(t k ),

[0201] Among them, ΔM(t k ) is time t k Feeding amount deviation, M(t k ) is the planned feeding amount;

[0202] The steps of conducting feedback analysis on the effect of the previous round of feeding and using the analysis results to optimize the next round of feeding plan also include:

[0203] Comprehensively consider the impact of environmental parameters on feeding behavior and define the environmental correction weight ω(t k ), the correction formula is:

[0204]

[0205] Among them, ω(t k ) is the environmental correction weight, α, β are the weight coefficients of dissolved oxygen and water temperature,

[0206] The corrected feeding bias is:

[0207] ΔS adj (t k )=ΔS(t k )+ω(t k );

[0208] Update the feeding model parameters based on the feedback results and optimize the next round of feeding plan. The optimization formula is:

[0209] M opt (t k+1 )=M(t k )+λ1ΔS adj (t k )-λ2ΔM(t k ),

[0210] Among them, M opt (t k+1 ) is time t k+1 The optimal feeding amount, λ1, λ2 are the adjustment coefficients of feeding deviation and feeding deviation,

[0211] Generate the next round of feeding plan:

[0212] P(t k+1 )={M opt (t k+1 ), T opt (t k+1 )}, where feeding time T opt (t k+1 ) is recalculated by analyzing the trend of feeding intensity changes, and the calculation formula is:

[0213]

[0214] Specifically, a comprehensive evaluation is conducted based on feeding intensity, feeding amount and environmental factors, and model parameters are adjusted dynamically to adapt to different environmental conditions.

[0215] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent feeding management system for aquaculture based on the Internet of Things, characterized by: include, IoT monitoring module, used to collect aquaculture data and extract fish characteristics data, including cameras and dissolved oxygen detectors, with built-in edge computing units; The local embedded computing module is used to receive the fish characteristic data transmitted by the IoT monitoring module, use lightweight algorithms to analyze the feeding status of the fish, and generate a feeding intensity and feeding demand model; it uses an embedded processor and is equipped with a lightweight target detection algorithm; A feeding control logic unit, including a control logic circuit and a calculation unit, is used to perform logic operations and feeding execution; In the feeding control logic, the feeding amount and feeding timing are calculated based on the feeding status and demand model output by the local embedded computing module and combined with the actual characteristic data; The feeder is used to receive instructions from the feeding control logic device, distribute feed according to the set feeding amount and time, and deliver it to the breeding area; The feeder includes a feed storage bin and a feeding execution device. The feeding execution device uses a rotating feeding tray and a track-type feeding device. The feeding data storage unit uses an embedded memory to store the feeding data of each feeding task, including feeding amount, time, feeding status analysis results and breeding data.

2. The aquaculture intelligent feeding management system based on the Internet of Things according to claim 1, characterized in that: The camera collects fish feeding behavior image data in real time to capture the movement trajectory, group density and feeding status of the fish; The dissolved oxygen detector monitors the dissolved oxygen level and water temperature in the fish pond where the fish are located.

3. The aquaculture intelligent feeding management system based on the Internet of Things as claimed in claim 2, characterized in that: The management method of the aquaculture intelligent feeding management system includes: Step S1, using a camera and a dissolved oxygen detector to collect aquaculture data, and extracting fish school characteristic data based on the aquaculture data; Step S2, the fish school characteristic data is transmitted to the local embedded computing module, and the lightweight target detection TinyYOLO algorithm is used to analyze the feeding status of the fish school to generate a feeding intensity and feeding requirement model; Step S3, taking the analysis result of the feeding state as input, combining the actual feeding demand and characteristic data, and using the feeding control logic to calculate the feeding amount and timing combination; Step S4, based on the calculation result of step S3, setting the automatic feeding instruction of the feeder, the feeding instruction matches the calculated feeding amount and feeding timing combination; Step S5, after completing the feeding task, the effect of the previous round of feeding is fed back and analyzed in combination with the changes in the feeding status and characteristic data of the fish school, and the analysis results are used to optimize the next round of feeding plan.

4. The aquaculture intelligent feeding management system based on the Internet of Things as claimed in claim 3, characterized in that: The breeding data include the movement trajectory, population density and feeding status of fish schools, and the dissolved oxygen and water temperature of the water quality.

5. The aquaculture intelligent feeding management system based on the Internet of Things as claimed in claim 4, characterized in that: The step of extracting fish swarm characteristic data based on breeding data is: Use the camera to collect time series images and generate a frame sequence {I1, I2, ..., I n }, where n is the total number of frames, I k Indicates time t k The image frames collected at the same time, k = 1, 2, ..., n, for each frame of image I k Grayscale processing is performed to convert color pixels into grayscale pixels. The grayscale formula is: G(x, y) = 0.299R(x, y) + 0.587G(x, y) + 0.114B(x, y), where n is the total number of frames in the time series, I k is time t k The image frame collected at , G(x, y) is the gray value of the pixel (x, y), R(x, y), G(x, y), B(x, y) are the values ​​of the pixel in the red, green, and blue channels, 0.299, 0.587, 0.114 are the weighting coefficients of the red, green, and blue channels to the gray value respectively; The optical flow method is used to calculate the motion vector u(x, y, t k ), the calculation formula is: Solve for the velocity vector: u(x, y, t k )=(u, v), where u(x, y, t k ) is the pixel point (x, y) at time t k The motion vector, u and v are the velocity components of the pixel along the x and y directions, is the spatial gradient of the image, is the temporal gradient of the image; Use background subtraction to extract the fish school area, count the number of pixels in the target area and calculate the school density. The extraction formula is: Among them, ρ is the fish density, B(x, y) is a binary mask, the fish pixels are 1, the background pixels are 0, R is the fish target area, and A R is the area of ​​the target region; The dissolved oxygen detector records water quality parameters at time intervals Δt to form a sequence of dissolved oxygen and water temperature, expressed as: Among them, W(t k ) is time t k The water quality characteristic vector at time , is time t k The dissolved oxygen concentration at water (t k ) is time t k The water temperature at that time, Δt is the sampling time interval of the dissolved oxygen detector; Merge feature data into time t k The eigenvector F(t k ), Among them, F(t k ) is time t k The fish school characteristic vector, u(x, y, t k ) is the fish school motion vector, ρ(t k ) is the population density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature.

6. The aquaculture intelligent feeding management system based on the Internet of Things as claimed in claim 5, characterized in that: The steps of using the lightweight target detection TinyYOLO algorithm to analyze the feeding status of fish schools and generate a feeding intensity and feeding requirement model are as follows: The fish school image feature F(t k ) is used as input for TinyYOLO to detect objects. Defining Inputs include: I ROI (t k ), is the time t k The image of the region of interest at that time, the cropped image range contains the feeding behavior of the fish school, H×W×c is the image size, which is height H, width W, and number of channels c=1; Using the TinyYOLO network structure, image I ROI (t k ) to extract features and generate a bounding box prediction set: in, is time t k is the bounding box set of the fish school target, b i =(x i ,y i , w i ,h i , p i ) is the feature of a single bounding box, where x i ,y i is the center coordinate of the bounding box, w i ,h i is the bounding box width and height, p i is the confidence level, m is the total number of bounding boxes detected; Based on the bounding box set Calculate the feeding intensity using the formula: Among them, S(t k ) is time t k The feeding intensity at that time, A ROI is the area of ​​the region of interest, p i is the bounding box b i The confidence level, A(b i )=w i ·h i is the bounding box b i area; Using feeding intensity S(t k ) and environmental parameters to generate a feeding demand model, the model formula is: Among them, N(t k ) is time t k The feeding requirement at that time, S(t k ) is time t k The intensity of feeding, is time t k The dissolved oxygen concentration at water (t k ) is time t k α, β, γ are the weight parameters of feeding intensity, dissolved oxygen and water temperature; The feeding intensity model S(t k ), describing the feeding activity of the fish at a specific moment, and the feeding demand model N(t k ), provide quantitative feeding amount requirements.

7. The aquaculture intelligent feeding management system based on the Internet of Things according to claim 6, characterized in that: The step of using the analysis result of the feeding state as input, combining the actual feeding demand and characteristic data, and using the feeding control logic to calculate the feeding amount and timing combination is as follows: Acquiring data from step S2 includes: Feeding intensity S(t k ), feeding requirement N(t k ) and the environmental characteristic vector E(t k ), in, is the dissolved oxygen concentration, θ water (t k ) is the water temperature; According to the feeding intensity S(t k ), environmental characteristics E(t k ), dynamic calculation time t k Feeding amount M(t k ), the calculation formula is: Among them, M(t k ) is time t k Feeding amount, N(t k ) is the feeding requirement, is the dissolved oxygen concentration, θ water (t k ) is water temperature, λ1, λ2, λ3 are weight coefficients of feeding demand, dissolved oxygen and water temperature; The feeding time is defined as the time window that meets the feeding needs of the fish. The feeding intensity change trend ΔS(t k ) to determine the optimal time, the calculation formula is: ΔS(t k )=S(t k )-S(t k-1 ), Trigger conditions for feeding time: If ΔS(t k )>θ S and Then T(t k ) = t k , Otherwise, T(t k ) is the delay, where T(t k ) is time t k The feeding trigger time, ΔS(t k ) is the change in feeding intensity, θ S is the feeding intensity threshold, is the dissolved oxygen concentration threshold; The final feeding plan includes the combination of feeding amount and time, and the formula is: P(t k )={M(t k ),T(t k )), Among them, P(t k ) is time t k The feeding plan, M(t k ) is the feeding amount, T(t k ) is the feeding time.

8. The aquaculture intelligent feeding management system based on the Internet of Things according to claim 7, characterized in that: The step of setting the automatic feeding instruction of the feeder based on the calculation result of step S3 is: The feeding plan P(t k ), including: feeding amount M(t k ) and feeding time T(t k ); Generate feeding instructions, feeding instructions C(t k ) including: feeding start time and feeding equipment operating parameters, The formula for generating feeding instructions is: C(t k )={t s , v f , d s }, where C(t k ) is time t k Feeding instruction, t s =T(t k ) is the start time of the feeder, is the feeding speed of the feeder, where M(t k ) is the feeding amount, τ d The duration of feeding is preset according to the specifications of the feeder. is the feeding coverage density, where r is the rotation range of the feeder, A t is the area of ​​the target feeding area; The feeding command is sent to the feeder execution device, including: Feeding controller, used to start or stop feed feeding and control feeding speed v f and duration τ d , The rotating throwing device is used to t and feeding coverage density d s , dynamically adjust the rotation speed; The feeder performs operations according to the received instructions and monitors the feeding process through sensors. The execution process is as follows: Among them, M act (t k ) is the actual feeding amount, t s is the feeding start time, τ d is the feeding duration, v f (t) is the actual material feeding speed; The feeder will feed the actual amount M act (t k ), the feeding coverage and time are fed back to the feeding data storage unit, and the feedback record is: R(t k )={M act (t k ), d s , t s }, where R(t k ) is time t k Feeding records, M act (t k ) is the actual feeding amount, d s is the feeding coverage density, t s The actual startup time.

9. The aquaculture intelligent feeding management system based on the Internet of Things according to claim 8, characterized in that: The step of combining the feeding status of the fish school with the changes in the characteristic data to conduct feedback analysis on the effect of the previous round of feeding, and using the analysis results to optimize the next round of feeding plan is as follows: Obtain the actual feeding record R(t k ) and feeding status data S(t k ), and combined with the fish school characteristic data F(t k ) and environmental data E(t k ), forming the input for feedback analysis: Among them, A(t k ) is time t k Feedback analysis input data, M act (t k ) is the actual feeding amount, S(t k ) is the feeding intensity, ρ(t k ) is the fish density, is the dissolved oxygen concentration, θ water (t k ) is the water temperature; Compare the actual feeding intensity S(t k ) and expected feeding intensity S exp (t k ) to evaluate the feeding effect and calculate the feeding deviation. The calculation formula is: ΔS(t k )=S(t k )-S exp (t k ), Among them, ΔS(t k ) is time t k Deviation of feeding intensity, S exp (t k ) is time t k The expected feeding intensity, The actual feeding amount deviation calculation formula is: ΔM(t k )=M act (t k )-M(t k ), Among them, ΔM(t k ) is time t k Feeding amount deviation, M(t k ) is the planned feeding amount.

10. The aquaculture intelligent feeding management system based on the Internet of Things according to claim 9, characterized in that: The step of performing feedback analysis on the effect of the previous round of feeding and using the analysis results to optimize the next round of feeding plan also includes: Comprehensively consider the impact of environmental parameters on feeding behavior and define the environmental correction weight ω(t k ), the correction formula is: Among them, ω(t k ) is the environmental correction weight, α, β are the weight coefficients of dissolved oxygen and water temperature, The corrected feeding bias is: ΔS adj (t k )=ΔS(t k )+ω(t k ); Update the feeding model parameters based on the feedback results and optimize the next round of feeding plan. The optimization formula is: M opt (t k+1 )=M(t k )+λ1ΔS adj (t k )-λ2ΔM(t k ), Among them, M opt (t k+1 ) is time t k+1 The optimal feeding amount, λ1, λ2 are the adjustment coefficients of feeding deviation and feeding deviation, Generate the next round of feeding plan: P(t k+1 )={M opt (t k+1 ), T opt (t k+1 )}, where feeding time T opt (t k+1 ) is recalculated by analyzing the trend of feeding intensity changes, and the calculation formula is:

Images