A method and system for decision-making regarding sheep feeding based on dynamic monitoring

Abstract

This invention relates to a method and system for dynamic monitoring-based decision-making in sheep feeding, comprising: acquiring basic parameters and physiological characteristic parameters of the sheep, environmental parameters and human factors during normal sheep farming, to obtain a first set; filtering and dimensionally equalizing the first set to obtain a second set; constructing an initial decision model based on the second set and training the initial decision model to obtain a decision model; acquiring a real-time set of human factors and a real-time set of environmental parameters, combining the first set and the decision model to obtain a decision dataset; determining the hunger state of the sheep based on the decision dataset, and feeding the sheep according to the hunger state. This invention predicts physiological characteristic parameters, then determines the hunger state, and finally combines these with the sheep's basic parameters to determine the feeding amount.

Description

Technical Field

[0001] This invention relates to the field of smart farming technology, and in particular to a method and system for decision-making regarding sheep feeding based on dynamic monitoring. Background Technology

[0002] With social development and rising living standards, people have gradually recognized the nutritional and health benefits of mutton, which is high in protein and low in fat. This has led to a significant increase in demand for mutton, making the mutton sheep farming industry a promising prospect. However, the amount and timing of feeding have become major challenges in mutton sheep farming. Insufficient feed intake prevents sheep from obtaining adequate nutrition and hinders their growth; conversely, excessive feed intake leads to feed waste, and overeating can cause discomfort, affecting their appetite for the next meal and even causing diarrhea. Summary of the Invention

[0003] In view of this, the present invention provides a method and system for decision-making on feeding of meat sheep based on dynamic monitoring. By predicting the physiological characteristic parameters of the animal, the hunger state is obtained, the feeding amount is obtained according to the hunger state and basic parameters, and then feeding is carried out according to the feeding amount.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A method for decision-making regarding sheep feeding based on dynamic monitoring, comprising:

[0006] The basic parameters and physiological characteristics of the sheep themselves, as well as the environmental parameters and human factors during the normal breeding process of the sheep, were obtained to form the first set.

[0007] The first set is filtered and processed to have the same dimensions to obtain the second set;

[0008] An initial decision model is constructed based on the second set, and the initial decision model is trained to obtain a decision model;

[0009] Obtain a real-time set of human factors and a real-time set of environmental parameters, and combine the first set with the decision model to obtain a decision dataset;

[0010] The hunger status of the sheep is obtained based on the decision dataset, and the sheep are fed according to the hunger status.

[0011] Preferably, each parameter in the first set includes dynamic, real-time, and continuous data;

[0012] The basic parameters include the height, weight, age, and body fat percentage of the sheep.

[0013] The environmental parameters include temperature, relative humidity, light intensity, air velocity, relevant gas concentration, dust concentration, atmospheric pressure, noise intensity, and farm location.

[0014] The relevant gas concentrations include oxygen concentration, carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration, and carbon monoxide concentration;

[0015] The human factors mentioned include feeding activities, vehicles passing through sheep pens, cleaning sheep pens, vaccination, catching, shearing, construction, aircraft noise, human conversation and transport of manure.

[0016] The feeding activities include feeding concentrated feed, feeding hay, and artificial feeding;

[0017] The physiological characteristics parameters include heart rate, body temperature, blood pressure, blood sugar, respiration, steps, three-dimensional acceleration, angular velocity, and vocalization;

[0018] The first set of data is collected through a sensor module, which includes an environmental sensor unit, a related flexible wearable sensor unit, and a monitoring unit.

[0019] The environmental sensor unit includes a hygrometer, a light intensity meter, an anemometer, a toxic gas detector, a dust concentration detector, a barometer, and a noise meter; the flexible wearable sensor includes a gyroscope, a temperature patch, a heart rate monitor, and a pedometer.

[0020] The first set is analyzed to obtain a hunger threshold; based on the hunger threshold and the decision dataset, the hunger state of the sheep is obtained.

[0021] Preferably, the step of filtering and dimensionless processing of the first set to obtain the second set specifically involves:

[0022] The data in the first set is checked for missing data. If there is missing data, the missing data is supplemented to obtain a supplementary dataset. The supplementary dataset includes a supplementary set of basic parameters of sheep, a supplementary set of training environment parameters, a supplementary set of training human factors, and a supplementary set of training physiological characteristic parameters.

[0023] The supplementary formula is as follows:

[0024]

[0025] in,

[0026]

[0027] λ j ,λ j ∈[0,1];1≤m≤5,1≤n≤5,m,n∈N + ;

[0028] In the formula: X kX represents the supplementary value at the k-th missing data point, m represents the m data points obtained from the same sensor monitoring at the k-th missing data point, and X represents the supplementary value at the k-th missing data point. i λ represents the i-th data value obtained from the forward pass. i X represents the weight of the i-th data value, n represents the n data points obtained from the same sensor monitoring at the k-th missing data point, and X j λ represents the j-th data value obtained from the previous iteration. j N represents the weight of the j-th data value. + Represents the set of positive integers;

[0029] Anomaly detection is performed on the data in the supplementary set of training physiological feature parameters. If the data exceeds a first preset threshold range, the data is defined as the first abnormal data. The supplementary set of training physiological feature parameters is traversed to obtain the first abnormal dataset. The first preset threshold range is greater than or equal to... and less than or equal to X min X represents the minimum value of each physiological characteristic parameter of a meat sheep under normal conditions. max This represents the maximum value of each physiological characteristic parameter for sheep under normal conditions.

[0030] When the ratio of the first abnormal dataset to the first data volume of the training physiological feature parameter supplementary set is less than a first predetermined ratio, the first abnormal data in the first abnormal dataset is replaced, and the replacement formula is as follows:

[0031]

[0032] In the formula: For X b The replacement value, X b Let X represent the b-th first outlier, q represent the length of the symmetrical interval spreading out from the node of the b-th first outlier, and X represent the length of the interval spreading out from the node of the b-th first outlier. i This represents the i-th normal data value within an interval of length q, and α represents the mean coefficient. When the first outlier is less than... When α∈[-0.3,0]; when the first outlier is greater than When α∈[0,0.3];

[0033] When the first data volume ratio is greater than or equal to the first set ratio, the acquisition process of the training physiological feature parameter set is checked. If the acquisition process is abnormal, the first abnormal dataset is discarded.

[0034] If the acquisition process is normal, the first abnormal dataset is judged. When the first abnormal data in the first abnormal dataset exceeds the second preset threshold range, the first abnormal data is defined as the second abnormal data. The first abnormal dataset is then traversed to obtain the second abnormal dataset. The second preset threshold range is greater than or equal to... and less than or equal to

[0035] When the ratio of the second abnormal dataset to the first abnormal dataset is less than a second set ratio, the second abnormal dataset is replaced; when the ratio of the second abnormal dataset is greater than or equal to the second set ratio, the second abnormal dataset is fitted to obtain a fitting function, and the average value of the second abnormal dataset is calculated.

[0036] If the fitting function exhibits a regular change, the second abnormal dataset is processed according to the following formula to obtain the first filtered dataset;

[0037]

[0038] in, For the processed second abnormal dataset, [X1,X2,…,X] p [X] represents the second abnormal dataset before processing. avg The average value of the second abnormal dataset; when When, the expression takes a negative sign, when When this happens, the expression takes the positive sign;

[0039] If the fitted function does not exhibit regular changes, the second abnormal dataset will be discarded.

[0040] For data X in the first filtered dataset c Make a judgment when or If the incident is caused by the passage of a feed truck or a loader, further investigation is needed to determine if shearing, cleaning the sheep pen, or vaccination occurred at the same time. If these actions occurred, no action is taken; otherwise, X will be... c The following formula is used for processing:

[0041]

[0042] in, For X c The value obtained after processing, X e X is the c-th data point in the first filtered dataset. c-1 For X c Previous data value, X c+1 For Xc The next data value, Ω i The values ​​for the human factors involved are r, where r represents the number of human factors involved, and ε is the difference coefficient.

[0043] Iterate through the first filtered dataset to obtain the second filtered dataset;

[0044] The second set is obtained by performing dimensionality processing on the second screening dataset, the supplementary set of basic parameters of the mutton sheep, and the supplementary set of training environment parameters, and by digitizing the supplementary set of training human factors.

[0045] Preferably, the step of performing dimensionality-matching processing on the second screening dataset, the supplementary set of basic parameters for sheep, and the supplementary set of training environment parameters, and digitizing the supplementary set of human factors in training to obtain the second set, specifically involves:

[0046] The second screening dataset, the supplementary set of basic parameters of mutton sheep, and the supplementary set of training environment parameters are processed to be of the same dimension to obtain the second screening data set of the same dimension, the supplementary set of basic parameters of mutton sheep set of the same dimension, and the supplementary set of training environment parameters set of the same dimension.

[0047] The training human factor supplement set is textual information. The training human factor supplement set is digitized to obtain the training human factor dataset. The second set includes the second screening data set of the same dimension, the meat sheep basic parameter supplement set of the same dimension, the training environment parameter supplement set of the same dimension, and the training human factor dataset.

[0048] The influence weight of each human factor is determined according to its degree of influence, and the factors are divided into those that are influencing the data and those that are not. If a factor is not influencing the data, it is marked as "1" and if it is not, it is marked as "0".

[0049] The classification is based on factors such as the distance of human factors from the sheepfold, the level of noise generated, whether it directly affects the sheep, and whether it involves feeding activities;

[0050]

[0051] Among them, Ω k a represents the data after the k-th text information has been digitized. i A represents the influencing factor of the i-th data in the c-th text information. i This represents the weight of the i-th data influencing factor, such as the distance from the sheepfold and the magnitude of the noise generated, as mentioned above; n indicates that there are a total of n data influencing factors; B j b represents the weight of the j-th non-influencing factor. j This indicates that the j-th factor is a non-influencing factor, and m indicates that there are a total of m non-influencing factors.

[0052] The data influencing factors include human factors such as the distance from the sheepfold and the magnitude of noise generated;

[0053] The factors that are considered as influencing factors include whether they directly affect the sheep themselves and whether they involve feeding activities; if so, the value is 1, otherwise it is 0.

[0054] Preferably, the step of constructing an initial decision model based on the second set and training the initial decision model to obtain a decision model includes:

[0055] The initial decision model is constructed based on the second set; the formula for calculating the number of computation nodes in the initial decision model is as follows:

[0056]

[0057] In the formula: ω2 is the number of operation nodes, ω1 is the number of input nodes, ω3 is the number of output nodes, A1 represents the influence factor of the number of input nodes on the number of operation nodes, A2 represents the influence factor of the number of output nodes on the number of operation nodes, and when the difference between ω1 and ω3 is greater than or equal to the set difference in the number of nodes, A1 = A2 = 1; σ is a given constant, σ = 5;

[0058] The basic parameters of the sheep, the training environment parameters, and the training human factors dataset are input into the initial decision model to obtain the training physiological characteristic parameter prediction set.

[0059] Based on the training physiological feature parameter prediction set and the second screening data set of the same dimension, the error value is obtained;

[0060] Determine the adaptive coefficients, and based on the error value and the adaptive coefficients, correct the threshold and weights of the initial decision model for iterative optimization until the error value is less than the set error value, thus obtaining the initial trained decision model; the formula for the adaptive coefficients is as follows:

[0061] γ(μ)=2 δ-1 γ(μ-1)

[0062]

[0063] In the formula: γ is the adaptive coefficient, μ represents the μ-th iteration, γ(μ) represents the value of the adaptive coefficient of the μ-th iteration, and e is the base of the natural logarithm function; when μ=1, γ(1)=1;

[0064] The accuracy of the initial training decision model is judged. If the accuracy is greater than or equal to the accuracy set value, the initial training decision model is used as the decision model. If the accuracy is less than the accuracy set value, the threshold and weights of the prediction model are corrected, and iterative optimization is continued until the accuracy is greater than or equal to the accuracy set value.

[0065] Preferably, the step of obtaining the hunger state of the sheep based on the decision dataset and feeding the sheep according to the hunger state includes:

[0066] The decision dataset is classified to obtain a classification dataset. Based on the classification dataset and the hunger threshold, the hunger state of the meat sheep is obtained.

[0067] The feeding amount is determined based on the hunger state of the sheep; feeding is then carried out based on the feeding amount.

[0068] Within a set time period after feeding the sheep, the collection frequency of the real-time human factor set and the real-time environmental parameter set is reduced.

[0069] Preferably, the step of obtaining the feeding amount based on the hunger state of the sheep, and feeding based on the feeding amount, includes:

[0070] The feeding amount is obtained based on the hunger state of the sheep and the supplementary set of basic parameters of the sheep; the feeding amount includes the amount of concentrate feed and the amount of hay.

[0071] The formula for calculating the amount of concentrated feed is as follows:

[0072]

[0073] The formula for calculating the amount of forage is as follows:

[0074]

[0075] In the formula: y i This is the i-th value in the supplementary set of basic parameters for meat sheep, where M is the amount of concentrated feed given to the meat sheep when it is in a state of complete starvation, and x. i Let m be the concentrate feed weight for the i-th value, m be the concentrate feed amount coefficient, N be the amount of forage for the sheep when it is in a state of complete starvation, and w be the feed amount for the sheep. i is the weight of the feed quantity for the i-th value, and n represents the feed quantity coefficient, which is obtained based on feed palatability, moisture content and nutrient composition.

[0076] A decision-making system for feeding meat sheep based on dynamic monitoring, comprising:

[0077] At the sensing end, the basic parameters and physiological characteristics of the sheep themselves, as well as the environmental parameters and human factors during the normal breeding process of the sheep, are obtained to form the first set; the real-time human factor set and the real-time environmental parameter set are also obtained.

[0078] In the cloud, the first set is filtered and processed to be of the same dimension to obtain a second set; an initial decision model is constructed based on the second set and trained to obtain a decision model; based on the real-time human factor set, the real-time environmental parameter set, and the first set, combined with the decision model, a decision dataset is obtained; the hunger state of the sheep is obtained based on the decision dataset; and the feeding amount is obtained based on the hunger state of the sheep.

[0079] The execution end feeds the animal according to the stated feeding amount.

[0080] The remote terminal displays the real-time human factor set, the real-time environmental parameter set, the decision dataset, the hunger status of the sheep, and the feeding amount.

[0081] The first set includes a set of basic parameters for meat sheep, a set of training environment parameters, a set of training human factors, and a set of training physiological characteristics parameters;

[0082] The sensing end includes:

[0083] A recording unit records a set of basic parameters for meat sheep; these basic parameters include height, weight, and age in months.

[0084] An environmental monitoring unit acquires the training environment parameter set and the real-time environment parameter set; the environmental parameters include temperature, humidity, light intensity, air velocity, relevant gas concentration, dust concentration, atmospheric pressure, noise, sheepfold area, and farm geographical location; the relevant gas concentrations include carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration, and carbon monoxide concentration.

[0085] The human monitoring unit acquires the real-time human factor set and the training human factor set; the human factors include feeding activities, cleaning sheep pens, vaccination, catching, shearing, construction, transporting manure, aircraft noise, and human conversation.

[0086] A flexible wearable sensor unit acquires the set of training physiological characteristic parameters; the physiological characteristic parameters include heart rate, body temperature, blood pressure, blood sugar, respiration, steps, acceleration, angular velocity, and vocalizations.

[0087] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0088] This invention relates to a method and system for dynamic monitoring-based decision-making in sheep feeding, comprising: acquiring basic parameters and physiological characteristic parameters of the sheep, environmental parameters and human factors during normal sheep farming, to obtain a first set; filtering and dimensionally equalizing the first set to obtain a second set; constructing an initial decision model based on the second set and training the initial decision model to obtain a decision model; acquiring a real-time set of human factors and a real-time set of environmental parameters, combining the first set and the decision model to obtain a decision dataset; determining the hunger state of the sheep based on the decision dataset, and feeding the sheep according to the hunger state. This invention predicts physiological characteristic parameters, then determines the hunger state, and finally combines these with the sheep's basic parameters to determine the feeding amount. Attached Figure Description

[0089] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0090] Figure 1 This is a flowchart of the dynamic monitoring-based decision-making method for meat sheep according to the present invention;

[0091] Figure 2 This is a structural diagram of the dynamic monitoring-based sheep feeding decision system of the present invention;

[0092] Figure 3 This is a simplified flowchart of the iterative optimization process of the decision model of this invention;

[0093] Figure 4 This is a flowchart of the iterative optimization process of the decision model of this invention;

[0094] Figure 5 This is a schematic diagram illustrating the principle of the dynamic monitoring-based decision-making system for meat sheep of the present invention.

[0095] Symbol explanation: 1-sensing end, 2-cloud, 3-execution end, 4-remote terminal. Detailed Implementation

[0096] 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.

[0097] The purpose of this invention is to provide an animal feeding method and system based on dynamic monitoring. By predicting the physiological characteristic parameters of the animal, the hunger state is obtained, the feeding amount is obtained based on the hunger state and basic parameters, and feeding is carried out according to the feeding amount.

[0098] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0099] Figure 1 This is a flowchart of the dynamic monitoring-based decision-making method for meat sheep according to the present invention. Figure 1 As shown, this invention provides a method for decision-making regarding sheep feeding based on dynamic monitoring, comprising:

[0100] Step S1: Obtain the basic parameters and physiological characteristics of the sheep, as well as the environmental parameters and human factors during the normal breeding process of the sheep, to obtain the first set.

[0101] The data includes the height, weight, age, and body fat percentage of the sheep.

[0102] The environmental parameters include temperature, relative humidity, light intensity, air velocity, relevant gas concentrations, dust concentration, atmospheric pressure, noise intensity, thunder, rain sound, and farm location; the relevant gas concentrations include oxygen concentration, carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration, and carbon monoxide concentration.

[0103] The human factors mentioned include feeding activities, vehicles passing through sheep pens, cleaning sheep pens, vaccination, catching, shearing, construction, aircraft noise, human conversations, and transporting manure.

[0104] The human factors mentioned above can be further subdivided in actual breeding processes: feeding activities include feeding concentrated feed, feeding hay, and manual feeding; these three activities have different effects on sheep, with sheep showing the most obvious reaction when fed concentrated feed; manual feeding refers to the process where feed is scattered outside the sheep's feeding area due to their eating activities, requiring manual shoveling of the feed back into the trough to reduce waste; in addition, different vehicles and personnel passing by also have different effects on sheep. During breeding, electric vehicles, concentrated feed trucks, hay trucks, feed trucks, and tricycles of various sizes will pass through the sheepfold, which can be categorized as passing through without stopping, stopping for a period of time, or passing by from outside the sheepfold; personnel passing by can also be categorized as ordinary personnel, staff, and feeders; and more human factors can be further subdivided according to the actual situation of the farm.

[0105] The aforementioned human factors can be eliminated through actual experiments, with some having minimal or no impact.

[0106] The physiological parameters include heart rate, body temperature, blood pressure, blood sugar, respiration, steps, three-dimensional acceleration, angular velocity, and vocalizations.

[0107] In actual experiments, the performance of experimental sheep was determined by applying different degrees of human interference. Environmental parameters such as temperature, relative humidity, noise, light, airflow and air quality affect the feeding activities of meat sheep. Human factors such as gathering feed, passing of feed trucks, cleaning of sheep pens and shearing are the main factors affecting the changes in the state of meat sheep. When meat sheep are in different states of hunger, their body temperature, blood sugar, bleating frequency and acceleration will show more obvious changes.

[0108] Among the human factors, the passage of feed collection and loading vehicles are both disruptive factors. If the sheep are not hungry, they will still show changes in their biological parameters to varying degrees when encountering the above situations. In this case, the judgment should be made in combination with the biological parameters over a period of time. When irregular and sudden situations such as shearing and cleaning the sheep pen occur, they will drastically affect the biological parameters of the sheep and also affect their hunger status. In this case, the changes in the vital signs of the sheep cannot be ignored.

[0109] The first set is acquired through a sensor module, which includes an environmental sensor unit, a related flexible wearable sensor unit, and a monitoring unit.

[0110] The environmental sensor unit includes a hygrometer, a light intensity meter, an anemometer, a toxic gas detector, a dust concentration detector, a barometer, and a noise meter; the flexible wearable sensor includes a gyroscope, a temperature patch, a heart rate monitor, and a pedometer. Human factors are obtained through recording or machine navigation.

[0111] Furthermore, each parameter in the first set includes dynamic, real-time, and continuous data.

[0112] The first set is analyzed to obtain a hunger threshold; based on the hunger threshold and the decision dataset, the hunger state of the sheep is obtained.

[0113] Step S2: Filter and perform dimensionless processing on the first set to obtain the second set.

[0114] Specifically, such as Figure 4 As shown, step S2 specifically includes:

[0115] The data in the first set is checked for missing data. If there is missing data, the missing data is supplemented to obtain a supplementary dataset. Missing data is defined as 0 data points. The supplementary dataset includes a supplementary set of basic parameters of sheep, a supplementary set of training environment parameters, a supplementary set of training human factors, and a supplementary set of training physiological characteristic parameters.

[0116] The supplementary formula is as follows:

[0117]

[0118] in,

[0119]

[0120] λ j ,λ j ∈[0,1];1≤m≤5,1≤n≤5,m,n∈N + ;

[0121] In the formula: X k X represents the supplementary value at the k-th missing data point, m represents the m data points obtained from the same sensor monitoring at the k-th missing data point, and X represents the supplementary value at the k-th missing data point. i λ represents the i-th data value obtained from the forward pass. i X represents the weight of the i-th data value, n represents the n data points obtained from the same sensor monitoring at the k-th missing data point, and X j λ represents the j-th data value obtained from the previous iteration. j N represents the weight of the j-th data value. + Let represent the set of positive integers; the inequality states that the weight of the forward data at a missing data point should be greater than the weight of the backward data, when The effect is best when it is at that time.

[0122] Anomaly detection is performed on the data in the supplementary set of training physiological feature parameters. If the data exceeds a first preset threshold range, the data is defined as the first abnormal data. The supplementary set of training physiological feature parameters is traversed to obtain the first abnormal dataset. The first preset threshold range is greater than or equal to... and less than or equal to X min X represents the minimum value of each physiological characteristic parameter of a meat sheep under normal conditions. max This represents the maximum value of each physiological characteristic parameter of a meat sheep under normal conditions.

[0123] When the ratio of the first abnormal dataset to the first data volume of the training physiological feature parameter supplementary set is less than a first predetermined ratio, the first abnormal data in the first abnormal dataset is replaced, and the replacement formula is as follows:

[0124]

[0125] In the formula: For X b The replacement value, X bLet X represent the b-th first outlier, q represent the length of the symmetrical interval spreading out from the node of the b-th first outlier, and X represent the length of the interval spreading out from the node of the b-th first outlier. i This represents the i-th normal data value within the interval length, and α represents the mean coefficient. When the first outlier is less than... When α∈[-0.3,0]; when the first outlier is greater than When α∈[0,0.3].

[0126] When the first data volume ratio is greater than or equal to the first set ratio, the acquisition process of the training physiological feature parameter set is checked. If the acquisition process is abnormal, the first abnormal dataset is discarded.

[0127] If the acquisition process is normal, the first abnormal dataset is judged. When the first abnormal data in the first abnormal dataset exceeds the second preset threshold range, the first abnormal data is defined as the second abnormal data. The first abnormal dataset is then traversed to obtain the second abnormal dataset. The second preset threshold range is greater than or equal to... and less than or equal to

[0128] When the ratio of the second abnormal dataset to the first abnormal dataset is less than a second set ratio, the second abnormal dataset is replaced. When the ratio of the second abnormal dataset is greater than or equal to the second set ratio, the second abnormal dataset is fitted to obtain a fitting function, and the average value of the second abnormal dataset is calculated.

[0129] If the fitting function exhibits a regular change, the second abnormal dataset is processed according to the following formula to obtain the first filtered dataset;

[0130]

[0131] in, For the processed second abnormal dataset, [X1,X2,…,X] p [X] represents the second abnormal dataset before processing. avg The average value of the second abnormal dataset; when When, the expression takes a negative sign, when When the expression is positive, the expression takes the positive sign.

[0132] If the fitted function does not exhibit regular changes, then the second abnormal dataset is discarded.

[0133] For data X in the first filtered dataset e Make a judgment when or If the incident is caused by the passage of a feed truck or a loader, further investigation is needed to determine if shearing, cleaning the sheep pen, or vaccination occurred at the same time. If these actions occurred, no action is taken; otherwise, X will be... e The following formula is used for processing:

[0134]

[0135] in, For X c The value obtained after processing, X e X is the c-th data point in the first filtered dataset. c-1 For X c Previous data value, X c+1 For X c The next data value, Ω i The values ​​for the human factors involved are r, where r represents the number of human factors involved, and ε is the difference coefficient.

[0136] Iterate through the first filtered dataset to obtain the second filtered dataset.

[0137] The second set is obtained by performing dimensionality processing on the second screening dataset, the supplementary set of basic parameters of the mutton sheep, and the supplementary set of training environment parameters, and by digitizing the supplementary set of training human factors.

[0138] The formula for processing the same dimensions in the supplementary set of basic parameters for meat sheep is as follows:

[0139]

[0140] In the formula: E ij e represents the result of processing the j-th data point of the i-th basic parameter in the supplementary set of basic parameters for mutton sheep to the same dimensions. j Let represent the j-th data point in the i-th basic parameter of the supplementary set of basic parameters for mutton sheep, and z represent the total number of data points for the i-th basic parameter. This represents the minimum value among z data points. This represents the largest data point among z data points.

[0141] The formula for processing the dimensionless parameters of the training environment supplementary set is as follows:

[0142]

[0143] Among them, M ij This represents the result of processing the j-th data point in the i-th environment parameter of the training environment parameter supplement set to the same dimensions, where s represents the total number of data points in the i-th environment parameter, and m represents the result of processing the j-th data point in the training environment parameter supplement set to the same dimensions. j This represents the j-th data point in the i-th environment parameter of the training environment parameter supplement set.

[0144] The formula for processing the second filtered dataset with the same dimensions is as follows:

[0145]

[0146]

[0147] Where, N ij This represents the result of processing the j-th data point of the i-th physiological characteristic parameter in the second screening dataset with the same dimensions, where n i Let represent the j-th data point of the i-th physiological feature parameter in the second screening dataset, d represent the total number of data points for the i-th physiological feature parameter, and Z represent the average value of the d data points.

[0148] The training human factor supplement set is textual information. The training human factor supplement set is digitized to obtain the training human factor dataset. The second set includes the second screening data set with the same dimensions, the sheep basic parameter supplement set with the same dimensions, the training environment parameter supplement set with the same dimensions, and the training human factor dataset.

[0149] The influence weight of each human factor is determined according to its degree of influence, and the factors are divided into those that are influencing the data and those that are not. If a factor is not influencing the data, it is defined as "1" and if it is not, it is "0".

[0150] The factors affecting human activity are categorized based on their distance from the sheepfold, the level of noise generated, whether they directly impact the sheep, and whether they involve feeding activities.

[0151]

[0152] Among them, Ω k a represents the data after the k-th text information has been digitized. i A represents the influencing factor of the i-th data in the c-th text information. i This represents the weight of the i-th data influencing factor, such as the distance from the sheepfold and the magnitude of the noise generated, as mentioned above; n indicates that there are a total of n data influencing factors; B j b represents the weight of the j-th non-influencing factor. j This indicates that the j-th factor is a non-influencing factor, and m indicates that there are a total of m non-influencing factors.

[0153] The data influencing factors include human factors such as the distance from the sheepfold and the magnitude of noise generated;

[0154] The factors that are considered as influencing factors include whether they directly affect the sheep themselves and whether they involve feeding activities; if so, the value is 1, otherwise it is 0.

[0155] Step S3: Construct an initial decision model based on the second set, and train the initial decision model to obtain a decision model.

[0156] Specifically, such as Figure 3 and Figure 4 As shown, step S3 includes:

[0157] The initial decision model is constructed based on the second set; the formula for calculating the number of computation nodes in the initial decision model is as follows:

[0158]

[0159] In the formula: ω2 is the number of operation nodes, ω1 is the number of input nodes, ω3 is the number of output nodes, A1 represents the influence factor of the number of input nodes on the number of operation nodes, A2 represents the influence factor of the number of output nodes on the number of operation nodes, and when the difference between ω1 and ω3 is greater than or equal to the set difference in the number of nodes, A1 = A2 = 1; σ is a given constant, σ = 5.

[0160] The basic parameters of the sheep, the training environment parameters, and the training human factors dataset are input into the initial decision model to obtain the training physiological characteristic parameter prediction set.

[0161] Calculate the cumulative amount for each computing node and obtain the output value for each computing node.

[0162] The formula for calculating the cumulative amount of a computing node is as follows:

[0163]

[0164] Where f i w represents the cumulative amount at the i-th operation node. ij Let x represent the weight of the j-th parameter with respect to the i-th operation node. ij σ1 represents the value corresponding to the j-th parameter, g represents the specific value assigned in the initial calculation, and g represents the number of data corresponding to the i-th operation node.

[0165] The numerical calculation formula for the output of the operation node is as follows:

[0166]

[0167] Where S i represents the value output by the i-th operation node, and e represents the base of the natural logarithm function.

[0168] Calculate the cumulative amount for each output node and obtain the value output by each output node.

[0169] The formula for calculating the cumulative amount of the output node is as follows:

[0170]

[0171] Where, m k w represents the cumulative amount at the k-th output node. ki β1 represents the weight of the i-th operation node relative to the k-th output node, β2 represents the specific value assigned in the initial calculation, and n represents that the k-th output node corresponds to n operation nodes.

[0172] The formula for calculating the numerical value output by the output node is as follows:

[0173]

[0174] Where M k This represents the value output by the k-th output node, and e represents the base of the natural logarithm function.

[0175] Based on the training physiological feature parameter prediction set and the second screening data set of the same dimensions, the error value is obtained.

[0176] Determine the adaptive coefficients, and based on the error value and the adaptive coefficients, correct the threshold and weights of the initial decision model for iterative optimization until the error value is less than the set error value, thus obtaining the initial trained decision model; the formula for the adaptive coefficients is as follows:

[0177] γ(μ)=2 δ-1 γ(μ-1)

[0178]

[0179] In the formula: γ is the adaptive coefficient, μ represents the μ-th iteration, γ(μ) represents the value of the adaptive coefficient of the μ-th iteration, and e is the base of the natural logarithm function; when μ=1, γ(1)=1.

[0180] The thresholds and weights of the initial decision model are corrected using the following formula:

[0181]

[0182]

[0183] Where, Δw ki Let Δβ2 be the weight correction amount of the i-th operation node relative to the k-th output node, Δβ2 be the threshold correction amount of the operation node relative to the output node, γ be the adaptive coefficient, and X be the weight correction amount of the i-th operation node relative to the k-th output node. k Let $\frac{ ... ki The influence factor, η2, represents the influence factor of the adaptive coefficient γ on Δβ2;

[0184] The accuracy of the initial training decision model is judged. If the accuracy is greater than or equal to the accuracy set value, the initial training decision model is used as the decision model. If the accuracy is less than the accuracy set value, the threshold and weights of the prediction model are corrected, and iterative optimization is continued until the accuracy is greater than or equal to the accuracy set value.

[0185] The computational threshold and weights of the initial trained decision model are corrected using the following formula:

[0186]

[0187]

[0188] Where, Δw ij Let Δβ1 be the weight correction amount of the j-th input node relative to the i-th operation node, γ be the threshold correction amount of the input node relative to the operation node, and X be the adaptive coefficient. k x is the actual value of the k-th output node. ij This represents the value that the j-th input node inputs to the i-th operation node, and ε1 represents the adaptive coefficient γ for Δw. ij The influence factor is ε2, which represents the influence factor of the adaptive coefficient γ on Δβ1.

[0189] Step S4: Obtain the real-time human factor set and the real-time environmental parameter set, and combine the first set and the decision model to obtain the decision dataset.

[0190] Verification experiments were conducted based on existing data, using both traditional methods and the method provided by this invention. The relevant data were input into the decision model, yielding experimental results as shown in Table 1. By comparing the mean, maximum, and minimum values ​​of the absolute values ​​of the relative errors, it can be concluded that the method of this invention achieves smaller errors and is suitable for practical production use.

[0191] Table 1 shows the decision dataset.

[0192]

[0193]

[0194] Step S5: Obtain the hunger status of the sheep based on the decision dataset, and feed the sheep according to the hunger status.

[0195] Specifically, step S5 includes: classifying the decision dataset to obtain a classification dataset, and obtaining the hunger state of the sheep based on the classification dataset and the hunger threshold.

[0196] The feeding amount is determined based on the hunger state of the sheep; feeding is then carried out based on the feeding amount.

[0197] Within a set time period after feeding the sheep, the collection frequency of the real-time human factor set and the real-time environmental parameter set is reduced.

[0198] The feeding amount is obtained based on the hunger state of the sheep and the supplementary set of basic parameters of the sheep; the feeding amount includes the amount of concentrate feed and the amount of hay.

[0199] The formula for calculating the amount of concentrated feed is as follows:

[0200]

[0201] The formula for calculating the amount of forage is as follows:

[0202]

[0203] In the formula: y i This is the i-th value in the supplementary set of basic parameters for meat sheep, where M is the amount of concentrated feed given to the meat sheep when it is in a state of complete starvation, and x. i Let m be the concentrate feed weight for the i-th value, m be the concentrate feed amount coefficient, N be the amount of forage for the sheep when it is in a state of complete starvation, and w be the feed amount for the sheep. i is the weight of the feed quantity for the i-th value, and n represents the feed quantity coefficient, which is obtained based on feed palatability, moisture content and nutrient composition.

[0204] like Figure 2 and Figure 5 As shown, the present invention provides a decision-making system for feeding meat sheep based on dynamic monitoring, comprising:

[0205] Sensing terminal 1 acquires the basic parameters and physiological characteristics of the sheep, as well as the environmental parameters and human factors during the normal breeding process of the sheep, to obtain the first set; and acquires the real-time human factor set and the real-time environmental parameter set.

[0206] In Cloud 2, the first set is filtered and processed to be of the same dimension to obtain a second set; an initial decision model is constructed based on the second set, and the initial decision model is trained to obtain a decision model; based on the real-time human factor set, the real-time environmental parameter set, and the first set, combined with the decision model, a decision dataset is obtained; the hunger state of the sheep is obtained based on the decision dataset; and the feeding amount is obtained according to the hunger state of the sheep.

[0207] Execution terminal 3 feeds the animal according to the feeding amount.

[0208] Remote terminal 4 can display the real-time human factor set, the real-time environmental parameter set, the decision dataset, the hunger status of the sheep, and the feeding amount.

[0209] The first set includes a set of basic parameters for meat sheep, a set of training environment parameters, a set of training human factors, and a set of training physiological characteristics parameters;

[0210] The sensing terminal 1 includes:

[0211] The recording unit records the basic parameters of the sheep, including height, weight, and age in months.

[0212] The environmental monitoring unit acquires the training environment parameter set and the real-time environment parameter set; the environmental parameters include temperature, humidity, light intensity, air velocity, relevant gas concentration, dust concentration, atmospheric pressure, noise, sheepfold area, and farm geographical location; the relevant gas concentrations include carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration, and carbon monoxide concentration.

[0213] The human monitoring unit acquires the real-time human factor set and the training human factor set; the human factors include feeding activities, cleaning sheep pens, vaccination, catching, shearing, construction, transporting manure, aircraft noise, and human conversation.

[0214] A flexible wearable sensor unit acquires the set of training physiological characteristic parameters; the physiological characteristic parameters include heart rate, body temperature, blood pressure, blood sugar, respiration, steps, acceleration, angular velocity, and vocalizations.

[0215] The cloud 2 includes:

[0216] A unit for processing equal dimensions filters and performs equal-dimensional processing on the first set to obtain a second set.

[0217] The machine training unit constructs an initial decision model based on the second set and trains the initial decision model to obtain a decision model.

[0218] The decision-making unit, based on the real-time human factor set, the real-time environmental parameter set, and the first set, and in conjunction with the decision-making model, obtains a decision dataset.

[0219] The processing unit obtains the hunger state of the sheep based on the decision dataset; and obtains the feeding amount based on the hunger state of the sheep.

[0220] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0221] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for feeding decision of mutton sheep based on dynamic monitoring, characterized in that, The method comprises the following steps: obtaining the basic parameters and physiological characteristic parameters of mutton sheep, the environmental parameters and human factors in the normal breeding process of mutton sheep, and obtaining a first set; screening and dimensionally processing the first set to obtain a second set; constructing an initial decision model based on the second set and training the initial decision model to obtain a decision model, comprising: constructing the initial decision model based on the second set; the number of operation nodes of the initial decision model is calculated according to the following formula: In the formula: is the number of operation nodes, is the number of input nodes, is the number of output nodes, represents an influence factor of the number of input nodes on the number of operation nodes, represents an influence factor of the number of output nodes on the number of operation nodes, when and are different by more than or equal to a node number setting difference, take ; is a given constant, = 5; inputting the mutton sheep basic parameter supplementary dimensionally set, the training environmental parameter supplementary dimensionally set and the training human factor data set into the initial decision model to obtain a training physiological characteristic parameter prediction set; obtaining an error value based on the training physiological characteristic parameter prediction set and the second screening data dimensionally set; determining an adaptive coefficient, correcting the threshold value and weight value of the initial decision model based on the error value and the adaptive coefficient for iterative optimization until the error value is less than an error setting value to obtain an initial training decision model; the formula of the adaptive coefficient is as follows: wherein: γ is an adaptive coefficient, representing the first iteration, representing the value of the adaptive coefficient for the first iteration, is the natural logarithm function base number; when , ; judging the accuracy of the initial training decision model; if the accuracy is greater than or equal to an accuracy setting value, the initial training decision model is taken as the decision model; if the accuracy is less than the accuracy setting value, the threshold value and weight value of the prediction model are corrected for continuous iterative optimization until the accuracy is greater than or equal to the accuracy setting value; obtaining a real-time human factor set and a real-time environmental parameter set, combining the first set and the decision model to obtain a decision data set; obtaining the hunger state of mutton sheep based on the decision data set and feeding according to the hunger state of mutton sheep.

2. The method according to claim 1, wherein, Each parameter in the first set comprises dynamic, real-time and continuous data; the basic parameters comprise the height, weight, age and body fat rate of mutton sheep; the environmental parameters comprise temperature, relative humidity, light intensity, air flow rate, related gas concentration, dust concentration, atmospheric pressure, noise intensity and farm geographical location; the related gas concentration comprises oxygen concentration, carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration and carbon monoxide concentration; the human factors comprise feeding activities, traffic passing through the sheep house, cleaning the sheep house, vaccination, catching, shearing, construction, airplane roaring sound, personnel conversation and transporting manure; the feeding activities comprise feeding concentrate, feeding forage and manual bunching; the physiological characteristic parameters comprise heart rate, body temperature, blood pressure, blood sugar, respiration, step count, three-dimensional acceleration, angular velocity and call; the first set is obtained by collecting through a sensor module, and the sensor module comprises an environmental sensor unit, a related flexible wearable sensor unit and a monitoring unit; the environmental sensor unit comprises a related hygrometer, a humidity meter, a light intensity measuring instrument, an anemometer, a toxic gas detector, a dust concentration detector, a barometer and a noise meter; the flexible wearable sensor comprises a gyroscope, a temperature patch, a heart rate meter and a pedometer; analyzing the first set to obtain a hunger threshold value; Based on the hunger threshold and the decision dataset, the hunger state of the mutton sheep is obtained.

3. The method according to claim 1, wherein, The first set is screened and processed in the same dimension to obtain a second set, specifically: The data in the first set is judged for missing data, and if there is data missing, the missing data is supplemented to obtain a supplemented dataset, which includes a mutton sheep basic parameter supplemented set, a training environment parameter supplemented set, a training artificial factor supplemented set, and a training physiological characteristic parameter supplemented set; the supplement formula is as follows: Wherein, wherein: represents a supplementary value at the th data missing point, represents the th data obtained by monitoring in the forward direction by the same sensor at the th data missing point, represents the th data value obtained in the forward direction, represents the weight of the th data value, represents the th data obtained by monitoring in the backward direction by the same sensor at the th data missing point, represents the th data value obtained in the backward direction, represents the weight of the th data value, represents a set of positive integers; The data in the training physiological characteristic parameter supplementary set are subjected to an abnormality judgment, and if the data exceeds a first set threshold range, the data is defined as first abnormal data, and the training physiological characteristic parameter supplementary set is traversed to obtain a first abnormal data set; the first set threshold range is greater than or equal to and less than or equal to ; ; ; is a minimum value of data corresponding to each physiological characteristic parameter of the mutton sheep in a normal state, is a maximum value of data corresponding to each physiological characteristic parameter of the mutton sheep in a normal state. When the first data quantity ratio of the first abnormal data set to the training physiological characteristic parameter supplemented set is less than a first set proportion, the first abnormal data in the first abnormal data set is replaced, and the replacement formula is as follows: , represents the bth first abnormal data, q represents the length of the interval of symmetry that spreads out at the node of the bth first abnormal data, represents the bth normal data value within the interval of length q, represents the mean coefficient, when the first abnormal data is less than , ; when the first abnormal data is greater than , ;​ When the first data quantity ratio is greater than or equal to the first set proportion, the acquisition process of the training physiological characteristic parameter set is checked, and if the acquisition process is abnormal, the first abnormal data set is discarded; If the acquisition process is normal, the first abnormal data set is judged. When the first abnormal data in the first abnormal data set exceeds a second set threshold range, the first abnormal data is defined as second abnormal data, the first abnormal data set is traversed, and a second abnormal data set is obtained. The second set threshold range is greater than or equal to and less than or equal to ; ; . When the second data quantity ratio of the second abnormal data set to the first abnormal data set is less than a second set proportion, the second abnormal data set is replaced, and when the second data quantity ratio is greater than or equal to the second set proportion, the second abnormal data set is fitted to obtain a fitting function and calculate the average value of the second abnormal data set; If the fitting function presents regular changes, the second abnormal data set is processed according to the following formula to obtain a first screening data set; wherein is the processed second set of outliers, is the second set of outliers before processing, is the mean of the second set of outliers; when the expression takes a negative sign, and when the expression takes a positive sign; If the fitting function does not present regular changes, the second abnormal data set is discarded; The data in the first screening data set are processed to determine if or occurred due to the passage of a loader or a loading wagon, and if, at the same time, shearing, cleaning the sheep pen, and vaccination occurred, no action is taken, and if they did not occur, the are processed by the following equation: wherein, is the value obtained after processing, is the cth data in the first screening data set, is the previous bit data value, is the next bit data value, is the value of the human factor that occurs, r represents that r kinds of human factors occur, is the difference coefficient; The first screening data set is traversed to obtain a second screening data set; The second screening data set, the mutton sheep basic parameter supplemented set, and the training environment parameter supplemented set are processed in the same dimension, and the training artificial factor supplemented set is dataized to obtain the second set.

4. The method according to claim 3, wherein, The second screening data set, the mutton sheep basic parameter supplemented set, and the training environment parameter supplemented set are processed in the same dimension, and the training artificial factor supplemented set is dataized to obtain the second set, specifically: The second screening data set, the mutton sheep basic parameter supplemented set, and the training environment parameter supplemented set are processed in the same dimension to obtain a second screening data same dimension set, a mutton sheep basic parameter supplemented same dimension set, and a training environment parameter supplemented same dimension set; The training artificial factor supplemented set is text information, and the training artificial factor supplemented set is dataized to obtain a training artificial factor dataset; the second set includes the second screening data same dimension set, the mutton sheep basic parameter supplemented same dimension set, the training environment parameter supplemented same dimension set, and the training artificial factor dataset; According to the influence degree, the influence weight of each artificial factor is determined, and the data influence factor and the yes-no influence factor are divided; if the yes-no influence factor is yes, it is defined as "1", and if it is no, it is defined as "0"; The distance of the artificial factor to the sheep shed, the size of the noise generated, and whether it directly acts on the mutton sheep itself or involves feeding activities are divided: in, Indicates the first Data after text information is digitized Indicates the first The first text message One data influencing factor, Indicates the first The weights of factors influencing the data, such as the distance from the sheepfold and the magnitude of noise generated, as mentioned above. Indicates a total of One data influencing factor; Indicates the first The weights of non-influencing factors Indicates the first One is a non-influencing factor, Indicates a total of One is a non-influencing factor; The data influence factor includes the distance of the artificial factor to the sheep shed and the size of the noise generated; The non-influencing factors include whether directly acting on the mutton sheep itself and whether involving feeding activities; if yes, 1, if not, 0.

5. The method according to claim 3, wherein the method is characterized by, The mutton sheep hunger state is obtained based on the decision data set, and feeding is performed according to the mutton sheep hunger state, including: The decision data set is classified to obtain a classification data set, and the mutton sheep hunger state is obtained based on the classification data set and the hunger threshold; According to the mutton sheep hunger state, a feeding amount is obtained; and feeding is performed based on the feeding amount; The acquisition frequency of the real-time human factor set and the real-time environmental parameter set is reduced within a set time length after the mutton sheep is fed.

6. The method according to claim 5, wherein, The feeding amount is obtained according to the mutton sheep hunger state; Feeding is performed based on the feeding amount, including: The feeding amount is obtained based on the mutton sheep hunger state and the mutton sheep basic parameter supplement set; the feeding amount includes a concentrate amount and a forage amount; The calculation formula of the concentrate amount is as follows: The calculation formula of the forage amount is as follows: In the formula: is a basic parameter of meat sheep, is a numerical value, is a concentrate amount when the meat sheep is in a complete starvation state, is a numerical value, is a concentrate weight value of the numerical value, is a concentrate amount coefficient; is a forage amount when the meat sheep is in a complete starvation state, is a numerical value, is a forage amount weight value of the numerical value, represents a forage amount coefficient, and the forage amount coefficient is obtained based on forage palatability, water content, and nutrient components.

7. A dynamic monitoring-based mutton sheep feeding decision system, characterized in that, Including: The perception end obtains the basic parameters and physiological characteristic parameters of the mutton sheep, the environmental parameters and human factors in the normal breeding process of the mutton sheep, and obtains a first set; The real-time human factor set and the real-time environmental parameter set are obtained; The cloud end filters and processes the first set to obtain a second set; An initial decision model is constructed based on the second set, and the initial decision model is trained to obtain a decision model, including: the initial decision model is constructed based on the second set; the calculation formula of the number of operation nodes of the initial decision model is as follows: In the formula: is the number of operation nodes, is the number of input nodes, is the number of output nodes, represents an influence factor of the number of input nodes on the number of operation nodes, represents an influence factor of the number of output nodes on the number of operation nodes, when and differ by more than or equal to a node number setting difference, take ; is a given constant, = 5; the initial decision model is input with the sheep basic parameter supplementary dimension set, the training environment parameter supplementary dimension set and the training human factor data set to obtain a training physiological characteristic parameter prediction set; based on the training physiological characteristic parameter prediction set and the second screening data dimension set, an error value is obtained; an adaptive coefficient is determined, the threshold and weight of the initial decision model are corrected based on the error value and the adaptive coefficient to perform iterative optimization until the error value is less than an error setting value, and an initial training decision model is obtained; the formula of the adaptive coefficient is as follows: wherein: γ is an adaptive coefficient, representing the first iteration, representing the first iteration of the adaptive coefficient, is a natural logarithm function base number; when , ; judging an accuracy of the initial training decision model, if the accuracy is greater than or equal to an accuracy setting value, taking the initial training decision model as the decision model; if the accuracy is less than the accuracy setting value, correcting a threshold value and a weight value of the prediction model, and continuing iteration optimization until the accuracy is greater than or equal to the accuracy setting value; obtaining a decision data set based on the real-time human factor set, the real-time environmental parameter set and the first set in combination with the decision model; obtaining a sheep hunger state based on the decision data set; and obtaining a feeding amount according to the sheep hunger state. The execution end performs feeding according to the feeding amount; The remote terminal displays the real-time human factor set, the real-time environmental parameter set, the decision data set, the mutton sheep hunger state and the feeding amount; The first set includes a mutton sheep basic parameter set, a training environmental parameter set, a training human factor set and a training physiological characteristic parameter set; The perception end includes: The recording unit records the mutton sheep basic parameter set; the basic parameters include height, weight and age; The environmental monitoring unit obtains the training environmental parameter set and the real-time environmental parameter set; the environmental parameters include temperature, humidity, light intensity, air flow rate, related gas concentration, dust concentration, atmospheric pressure, noise, sheep shed area and farm geographical location; the related gas concentration includes carbon dioxide concentration, hydrogen sulfide concentration, ammonia concentration and carbon monoxide concentration; The human monitoring unit obtains the real-time human factor set and the training human factor set; the human factors include feeding activities, cleaning the sheep pen, vaccination, capture, shearing, construction, transporting manure, airplane roaring sound and personnel conversation; The flexible wearable sensor unit obtains the training physiological characteristic parameter set; the physiological characteristic parameters include heart rate, body temperature, blood pressure, blood sugar, respiration, step count, acceleration, angular velocity and call sound.

Images