Mutton sheep state sensing method and system in feeding experiment
通过获取和分析肉羊的多种参数数据,利用最大间隔分类模型建立状态感知模型,解决了饲养实验中肉羊状态难以精确获取的问题,实现了对肉羊状态的精确监测和预测,提高了实验的准确 Sexual and economic benefits.
Patent Information
- Application Number
- CN202411075735.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-07
- Publication Date
- 2024-11-22
AI Technical Summary
In feeding experiments, it is difficult for the prior art to accurately obtain the status information of meat sheep, which affects the accuracy and economic benefits of experimental results.
By obtaining the basic data, biochemical parameters, vital sign parameters, biological apparent parameters and environmental parameter data of the meat sheep, data feature extraction and classification are carried out, a state-aware data set is constructed, and the maximum interval classification model is used for training and verification, and the state of the meat sheep is established is established by establishing the meat sheep state Perceptual model to monitor and predict the status of meat sheep in real time.
More accurate prediction of the status of meat sheep is achieved, helping experimenters to perceive and respond to abnormal situations in a timely manner, ensure the normal progress of the experiment, and improve breeding efficiency and economic benefits.
Smart Images

Figure CN119004176A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of precision and intelligent breeding technology, and in particular to a method and system for sensing the state of mutton sheep in a breeding experiment. Background Art
[0002] Feeding experiments are used to evaluate the effects of different feeds and feeding conditions on the growth, health and output of farm animals. The implementation of the experiment will help optimize breeding conditions, improve the production efficiency and health level of mutton sheep, thereby promoting the healthy development of the mutton industry, while also meeting people's demand for high-quality meat and achieving a virtuous interaction between agricultural production and people's lives. Through feeding experiments, we can gain an in-depth understanding of the metabolic response of mutton sheep to the components of different feeds, thereby optimizing feed formulas, ensuring that their nutritional needs are fully met, and improving growth performance and health. In feeding experiments, the state of mutton sheep directly affects their physiological activities. Sick or stressed mutton sheep may produce abnormal physiological phenomena, thereby affecting the accuracy of the experimental results. However, there are some limitations in existing technologies, and it is impossible to accurately obtain information on the state of animals in feeding experiments.
[0003] In summary, the research and development of methods and systems for sensing the status of mutton sheep during feeding experiments has important scientific and practical significance. It can fill the technical gaps in existing feeding experiments, improve animal production efficiency, and promote the development of related industries. Summary of the invention
[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method and system for sensing the status of mutton sheep in a feeding experiment. It aims to solve the problem of unclear individual status of mutton sheep in feeding experiments, ensure that the feeding experiments can be carried out and completed under normal and controllable conditions, and avoid economic losses caused by experimental failures.
[0005] To achieve the above object, the present invention provides the following solutions:
[0006] A method for sensing the state of mutton sheep in a feeding experiment, comprising:
[0007] Obtaining basic information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and inputting the basic information into a preset knowledge base;
[0008] Performing window processing and data feature extraction on the vital sign parameter data to obtain vital sign parameter feature data, and constructing a sheep state perception data set based on the vital sign parameter feature data, biochemical parameter data, biological appearance parameter data and environmental parameter data;
[0009] Determine and classify the sheep state perception data set according to a threshold determination function, and define labels for data in the sheep state perception data set according to the determination and classification result to construct a sheep state perception data training set;
[0010] Input the mutton sheep state perception data training set into the preset mutton sheep state perception model for training, verification and testing to obtain a constructed mutton sheep state perception model;
[0011] Based on obtaining new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target sheep in the feeding experiment, a new sheep state perception data set is constructed, and the new sheep state perception data set is input into the constructed sheep state perception model to obtain the state of the target sheep in the feeding experiment.
[0012] Preferably, the basic data information includes the number, strain, gender, age, health status and previous treatment of the target mutton sheep; the previous treatment refers to the treatment of the target mutton sheep before the feeding experiment begins.
[0013] Preferably, the biochemical parameter data include blood oxygen saturation, blood glucose concentration and blood lactate concentration; the vital sign parameter data include sound frequency change data, body temperature change data, heart rate change data and respiratory rate change data; the biological apparent parameter data include body weight, food intake and excretion; the environmental parameter data include noise, temperature, humidity, carbon dioxide concentration, oxygen concentration and carbon monoxide concentration.
[0014] Preferably, the vital sign parameter data is subjected to window processing and data feature extraction to obtain vital sign parameter feature data, and a sheep state perception data set is constructed based on the vital sign parameter feature data, biochemical parameter data, biological appearance parameter data and environmental parameter data, including:
[0015] Data features are extracted from the vital signs parameter data according to a preset vital signs parameter data feature extraction algorithm to obtain vital signs parameter feature data; a mutton sheep state perception data matrix is constructed according to the biochemical parameter data, the biological appearance parameter data, the vital signs parameter feature data and the environmental parameter data, and each of the mutton sheep state perception data matrices is assembled to obtain the mutton sheep state perception data set.
[0016] Preferably, the feature calculation formula in the data feature extraction is:
[0017]
[0018] Where: x irepresents the i-th data point in any windowed vital sign parameter data; N represents the total number of windowed vital sign parameter data; P(k) represents the power spectral density of the frequency spectrum of any windowed vital sign parameter data; f(n) represents the amplitude of any windowed vital sign parameter data; x(n) represents the frequency of any windowed vital sign parameter data; p represents the exponential parameter; X(k) represents the complex amplitude spectrum of any windowed vital sign parameter data in the frequency domain; μ4 represents the fourth-order central moment of the frequency spectrum of any windowed vital sign parameter data; μ2 represents the second-order central moment of the frequency spectrum of any windowed vital sign parameter data.
[0019] Preferably, the expression of the sheep state perception data matrix includes:
[0020]
[0021] Where P bio , P org , P phy , P env The sheep state perception data matrix respectively represents the parameter matrix composed of the biochemical parameter data, the biological appearance parameter data, the vital sign parameter characteristic data and the environmental parameter data; bio, org, phy and env respectively represent the biochemical parameter data, the biological appearance parameter data, the vital sign parameter characteristic data and the environmental parameter data; represents the i-th row data vector in the k-parameter data, i=1,2,3,…,m; m represents The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; j = 1, 2, 3, 4; T represents matrix transpose.
[0022] Preferably, the formula of the threshold determination function includes:
[0023]
[0024] In the formula, is the preset threshold; is the data vector of row i in the k parameter data matrix of the sheep state perception data, i=1,2,3,…,m; m represents The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; represents the threshold range parameter; express The weight j parameter of express The weight l parameter; express The data in the mutton state perception data matrix with c rows and q columns.
[0025] Preferably, label definition is performed on the data in the sheep state perception data set according to the determination and classification result, including:
[0026] when In When it is within the threshold range, the data label is classified as normal; when In If it is not within the threshold range, the data label is classified as an abnormal state.
[0027] Preferably, the sheep state perception data training set is input into a preset sheep state perception model for training, verification and testing to obtain a constructed sheep state perception model, including:
[0028] Dividing the sheep state perception data training set into a training set, a validation set and a test set according to a preset ratio;
[0029] The training set, validation set and test set are respectively input into the preset mutton state perception model for training, validation and testing, and the optimal decision boundary is determined to obtain the constructed mutton state perception model; the preset mutton state perception model is constructed based on the maximum interval classification model; the maximum interval classification model is composed of an input space, a mapping space, a decision boundary and an output space; the maximum interval classification model is used to map the mutton state perception data training set to different category labels; the input space is the space of the original data, and each feature vector is a sample as a point in the input space; the mapping space is constructed based on a mapping function; the mapping space is used to map the original data in the input space to a high-dimensional space to convert the original indivisible data into separable data; the decision boundary is a transcendental curve / plane that divides different categories in the mapping space to maximize the distance interval between different categories; the output space outputs the results classified by the decision boundary; the formula of the mapping function includes: Where: K p represents the mapping function; is the data vector of row i in the k parameter data matrix of the sheep state perception data, i=1,2,3,…,m; m represents The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; ω v represents the weight coefficient of the mapping subfunction (v=1,2,3,4); T represents the matrix transpose; θ and δ represent the slope coefficient of the mapping sub-function respectively; ε represents the intercept coefficient of the mapping sub-function; r represents the intercept coefficient of the mapping sub-function; the formula of the decision boundary definition function includes: Where: represents the decision boundary definition function; b * represents the bias coefficient of the decision boundary definition function; N represents the number of sheep state perception data vectors; Represents the sheep state perception data vector The category label of u Represents the sheep state perception data vector Lagrange multiplier; C n represents a correction parameter, which is used to control the complexity of the sheep state perception model; the correction parameter C n The update iterative process is as follows: n =C n-1 ·m n +c n ; Where: m n 、c n represents the randomly generated update parameter; C n Indicates the previous step value C n-1 Multiply by the update factor m n Add the update coefficient c n The result; the formula of the maximum interval definition function includes: Where: F Mar represents the maximum interval definition function; b represents the bias coefficient of the maximum interval definition function.
[0030] Preferably, based on acquiring new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment and constructing a new mutton sheep state perception data set, and inputting the new mutton sheep state perception data set into the constructed mutton sheep state perception model, the state of the target mutton sheep in the feeding experiment is obtained, including:
[0031] The biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data in the mutton sheep state perception data set are normalized, and the weight coefficients of different data are determined according to the contribution of the data; after the mutton sheep state perception data set is normalized, a weight matrix is constructed in combination with the weight coefficients, and the weight matrix is input into the constructed mutton sheep state perception model to obtain an output category;
[0032] The calculation formula for the normalization process includes: Where: represents the normalized sheep state perception data vector; is the data vector of row i in the k parameter data matrix of the sheep state perception data, i=1,2,3,…,m; m represents The total number of biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; N represents the number of mutton state perception data vectors; the weight coefficient calculation formula includes: Where: Indicates P k The weight coefficient of the parameter matrix; the weight input matrix calculation formula includes: Where: W input represents the weight matrix; P bio , P org , P phy , P env Respectively represent the sheep state perception data matrix composed of the biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data; the formula of the output category includes: Where: R i represents the output result of category i; A, B, C, D, E respectively represent the normal state, disease state, stress state, panic state and death state of the target mutton sheep; represents the decision space of category i consisting of the decision boundary;
[0033] If the sheep are identified as abnormal during the feeding experiment, the knowledge base will be called up and personalized decisions will be made based on the rich experience of the experimenters;
[0034] A sheep state perception system in a feeding experiment, comprising:
[0035] A data acquisition module is used to acquire basic information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and input the basic information into a preset knowledge base;
[0036] A perception data construction module is used to extract data features from the vital sign parameter data to obtain vital sign parameter characteristic data, and to construct a sheep state perception data set based on the biological appearance parameter data, the vital sign parameter characteristic data, the biochemical parameter data and the environmental parameter data;
[0037] A training data construction module is used to determine and classify the sheep state perception data set according to a threshold determination function, and to define labels for the data in the sheep state perception data set according to the determination and classification result, so as to construct a sheep state perception data training set;
[0038] A model training module is used to input the sheep state perception data training set into a preset sheep state perception model for training, verification and testing to obtain a constructed sheep state perception model;
[0039] The mutton sheep state perception module is used to input the mutton sheep state perception data set into the constructed mutton sheep state perception model to obtain the state of the target mutton sheep in the feeding experiment.
[0040] The present invention discloses the following technical effects:
[0041] The present invention provides a method and system for sensing the state of mutton sheep in a feeding experiment. The method comprises: obtaining basic data information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and inputting the basic data information into a preset knowledge base; respectively extracting data features of the vital sign parameter data to obtain vital sign parameter feature data, and constructing a mutton sheep state sensing data set according to the biological appearance parameter data, vital sign parameter feature data, biochemical parameter data and environmental parameter data; and determining the mutton sheep state sensing data according to a threshold judgment function. The present invention can make a more accurate prediction of the state of the individual mutton sheep, help the experimenter perceive the state of the mutton sheep in the experiment, and ensure the normal progress of the mutton sheep breeding experiment. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0043] Figure 1 A flow chart of a method provided by an embodiment of the present invention;
[0044] Figure 2 A schematic diagram of a technical route provided for an embodiment of the present invention;
[0045] Figure 3 A schematic diagram of a specific flow chart of a method provided by an embodiment of the present invention;
[0046] Figure 4 A schematic diagram of the system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0049] Figure 1 A flow chart of a method provided by an embodiment of the present invention, such as Figure 1 As shown, the present invention provides a method for sensing the state of mutton sheep in a feeding experiment, comprising:
[0050] Step 100: Obtain basic information of the target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and input the basic information into a preset knowledge base;
[0051] Step 200: Window processing and data feature extraction are performed on the vital sign parameter data to obtain vital sign parameter feature data, and a sheep state perception data set is constructed based on the biological appearance parameter data, vital sign parameter feature data, biochemical parameter data, and environmental parameter data;
[0052] Step 300: classifying the sheep state perception data set according to the threshold judgment function, and labeling the data in the sheep state perception data set according to the judgment and classification result to construct a sheep state perception data training set;
[0053] Step 400: inputting the sheep state perception data training set into a preset sheep state perception model for training, verification and testing to obtain a constructed sheep state perception model;
[0054] Step 500: Based on the above steps, new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target sheep in the feeding experiment are obtained and a new sheep state perception data set is constructed; the new sheep state perception data set is input into the constructed sheep state perception model to obtain the state of the target sheep in the feeding experiment.
[0055] Figure 2 A schematic diagram of a technical route provided for an embodiment of the present invention; Figure 3 Schematic diagram of the specific process of the method for sensing the state of mutton sheep in the feeding experiment of the present invention; Figure 2 and Figure 3As shown, an embodiment of the present invention provides a method for sensing the state of mutton sheep in a feeding experiment, comprising:
[0056] S201, registering basic information of target mutton sheep in the feeding experiment, and inputting the basic information into a knowledge base; the knowledge base is a collection of knowledge that stores, organizes and manages knowledge; the basic information includes mutton sheep number, strain, gender, age, health status and pre-treatment; the pre-treatment refers to the treatment of mutton sheep before the feeding experiment begins, such as dietary restrictions, drug treatment, etc.;
[0057] The knowledge base is a collection of information used to store, organize and manage knowledge and data; the knowledge base includes structured and unstructured information, making the information easy to access and use to support decision-making, problem solving and knowledge sharing; the knowledge base can improve the efficiency and innovation ability of an organization and promote the transfer and accumulation of knowledge;
[0058] The structured information is data stored in a clear data format definition, including but not limited to tables and texts; the unstructured information is data without a clear data format definition, including but not limited to images, audio and video;
[0059] S202, dynamically, in real time and continuously acquiring biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the mutton sheep during the feeding experiment; performing window processing on the mutton sheep vital sign parameter data and extracting data features in combination with a vital sign parameter data feature extraction algorithm to acquire vital sign parameter feature data, and constructing a mutton sheep state perception data set according to the biochemical parameter data, biological appearance parameter data, vital sign parameter feature data and environmental parameter data;
[0060] The biochemical parameter data include but are not limited to blood oxygen saturation, blood glucose concentration and blood lactate concentration; the vital sign parameter data include but are not limited to sound frequency change data, body temperature change data, heart rate change data and respiratory rate change data; the biological apparent parameter data include but are not limited to body weight, food intake and excretion; the environmental parameter data include but are not limited to noise, temperature, humidity, carbon dioxide concentration, oxygen concentration, carbon monoxide concentration and atmospheric pressure;
[0061] Specifically, the biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the mutton sheep in the feeding experiment are dynamically, real-time and continuously acquired during the experiment through a multi-scale sensor integration unit; the integration of the multi-scale sensor integration unit includes but is not limited to a blood oxygen saturation monitoring sensor, a sound monitoring sensor, a weight monitoring sensor and a noise monitoring sensor;
[0062] The operation flow of the vital sign parameter data feature extraction algorithm is as follows: inputting the vital sign parameter data of the mutton sheep; performing window processing on the vital sign parameter data; processing the windowed vital sign parameter data according to a given vital sign parameter data feature extraction algorithm; outputting the extracted data features and saving them;
[0063] The feature calculation formula for data feature extraction is as follows:
[0064] Where: x i represents the i-th data point in any windowed vital sign parameter data; N represents the total number of windowed vital sign parameter data; P(k) represents the power spectral density of the frequency spectrum of any windowed vital sign parameter data; N represents the total number of windowed vital sign parameter data; f(n) represents the amplitude of any windowed vital sign parameter data; x(n) represents the frequency of any windowed vital sign parameter data; p represents the exponential parameter; X(k) represents the complex amplitude spectrum of any windowed vital sign parameter data in the frequency domain; μ4 represents the fourth-order central moment of the frequency spectrum of any windowed vital sign parameter data; μ2 represents the second-order central moment of the frequency spectrum of any windowed vital sign parameter data;
[0065] Preferably, there is an optional sheep state perception data set composition scheme: a sheep state perception data matrix can be constructed, which includes biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data; a data vector of the sheep state perception data matrix can describe the sheep state information at a certain time point; a plurality of sheep state perception data matrices are assembled to finally form a sheep state perception data set;
[0066] Preferably, the sheep state perception data matrix is in the following form:
[0067]
[0068] Where: P bio , P org , P phy , P env The data matrices of sheep state perception respectively represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; bio, org, phy and env respectively represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; (i=1,2,3,…,m; m represents The total number of biochemical parameter data, biological epigenetic parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; j = 1, 2, 3, 4) represents the data vector of the ith row in the k parameter data; T represents the linear operation of matrix transposition, which is used to interchange the rows and columns of the matrix, that is, convert the rows of the matrix into columns and vice versa;
[0069] S203, classifying the data according to the sheep state perception data set in combination with the threshold judgment function; labeling the data in the sheep state perception data set according to the judgment and classification results, and constructing a sheep state perception data training set; the classification results include normal state and abnormal state;
[0070] The formula of the threshold determination function is as follows:
[0071]
[0072] Where: is the threshold value; represents the data vector of row i in the k parameter data matrix of the sheep state perception data (i = 1, 2, 3, ..., m; m represents The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively;); represents the threshold range parameter; express The weight j parameter; express The weight l parameter; express The sheep state perception data of row c and column q in the middle; preferably, when In When it is within the threshold range, the data label is defined as normal; when In When it is not within the threshold range, the data label is defined as an abnormal state;
[0073] Preferably, in order to better illustrate this embodiment, Table 2 describes the normal state and some abnormal states of the mutton sheep in the feeding experiment in words; it is worth noting that the abnormal states described below are not all abnormal states of the mutton sheep in the feeding experiment. In the actual application process, the remaining abnormal states are also within the scope of consideration of the present invention;
[0074] Table 2
[0075]
[0076]
[0077] Through the above operations, a sheep state perception data training set is constructed; the sheep state perception data training set includes but is not limited to a normal state data training set, a disease state data training set, a stress state data training set, a panic state data training set and a death state data training set;
[0078] Preferably, the data defined as normal by the label is input into the knowledge base to ensure that the information in the knowledge base is accurate and reliable;
[0079] S204, dividing the sheep state perception data training set into a training set, a validation set and a test set according to a certain ratio; inputting the training set, the validation set and the test set into a preset sheep state perception model for training, validation and testing respectively, determining the optimal decision boundary, and finally obtaining a constructed sheep state perception model; the preset sheep state perception model is constructed based on a maximum margin classification model;
[0080] The maximum margin classification model is mainly composed of an input space, a mapping space, a decision boundary and an output space; the maximum margin classification model is an intelligent system that can map the sheep state perception data to different category labels;
[0081] The input space is the space of original data, and each feature vector is a sample as a point in the input space; the mapping space is constructed based on the mapping function, mapping the original data in the input space to a higher-dimensional space, turning the original indivisible data into separable data, and providing a basis for maximum interval classification; the decision boundary is a transcendental curve / plane that divides different categories in the mapping space, so that the distance interval between different categories is maximized; the output space outputs the results classified by the decision boundary;
[0082] The formula of the mapping function is as follows:
[0084]
[0085] Where: K p represents a mapping function; represents the data vector of row i in the k parameter data matrix of the sheep state perception data (i = 1, 2, 3, ..., m; m represents The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data in k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively); ω v represents the weight coefficient of the mapping subfunction (v=1, 2, 3, 4); T represents the linear operation of matrix transposition, which is used to interchange the rows and columns of the matrix, that is, convert the rows of the matrix into columns and vice versa; θ,δ represent the slope coefficients of the mapping sub-function; ε represents the intercept coefficient of the mapping sub-function; r represents the intercept coefficient of the mapping sub-function;
[0086] The formula of the decision boundary definition function is as follows:
[0087]
[0088] Where: f boundary represents the decision boundary definition function; b * represents the bias coefficient of the decision boundary definition function; N represents the number of sheep state perception data vectors; Represents the sheep state perception data vector The category label of u Represents the sheep state perception data vector Lagrange multiplier; K p represents the mapping function; C n represents a correction parameter, which is used to control the complexity of the sheep state perception model; the correction parameter C n The update iterative process is as follows: n =C n-1 ·m n +c n ; Where: m n 、c n represents the randomly generated update parameter; C n Indicates the previous step value C n-1 Multiply by the update factor m n Add the update coefficient c n Result;
[0089] The formula of the maximum interval definition function is as follows:
[0090]
[0091] Where: F Mar represents the maximum interval definition function; α u Represents the sheep state perception data vector Lagrange multiplier; N represents the number of sheep state perception data vectors; Represents the sheep state perception data vector The category label of ; b represents the bias coefficient of the maximum interval definition function;
[0092] Preferably, the mapping function of the maximum interval classification model has a weight coefficient ω in the mapping sub-function v Under the influence of (v=1, 2, 3, 4), there are four simple mapping function modes; in order to better explain the embodiments of the present invention, these four simple mapping functions are briefly described below;
[0093] The K1 does not need to introduce additional transformations or mappings, and directly calculates the inner product in the original mapping space; if the K1 is used as the mapping function of the maximum margin classification model, the decision boundary definition function of the maximum margin classification model is: Where N represents the number of sheep state perception data vectors; α u is the Lagrange multiplier corresponding to the sheep state perception data vector; Represents the sheep state perception data vector The category label of ; b1 represents the bias coefficient of the maximum interval definition function of K1 mapping;
[0094] The K2 involves nonlinear mapping of the feature space, and uses K2 to calculate the similarity between two samples in the mapping space; if the K2 is used as the mapping function of the maximum margin classification model, the decision boundary definition function of the maximum margin classification model is: Where N represents the number of sheep state perception data vectors; α u is the Lagrange multiplier corresponding to the sheep state perception data vector; Represents the sheep state perception data vector The category label; b2 represents the bias coefficient of the maximum interval definition function of K2 mapping; ε represents the intercept coefficient of the mapping subfunction; represents the slope coefficient of the mapping sub-function; r represents the intercept coefficient of the polynomial mapping function;
[0095] K3 involves nonlinear mapping of the mapping space and uses K3 to calculate the similarity between two samples in the mapping space; if the K bell As the mapping function of the maximum margin classification model, the decision boundary definition function of the maximum margin classification model is: Where N represents the number of sheep state perception data vectors; α u is the Lagrange multiplier corresponding to the sheep state perception data vector; Represents the sheep state perception data vector The category label of ; b3 represents the bias coefficient of the K3 maximum interval definition function; θ represents the slope coefficient of the mapping subfunction;
[0096] K4 maps the original data samples of the input space to a high-dimensional space, making the data linearly separable in the space; if the K4 is used as the mapping function of the maximum margin classification model, the decision boundary definition function of the maximum margin classification model is: Where N represents the number of sheep state perception data vectors; α u is the Lagrange multiplier corresponding to the sheep state perception data vector; Represents the sheep state perception data vector The category label; b4 represents the bias coefficient of the maximum interval definition function of K4; δ represents the slope coefficient of the mapping subfunction; r represents the intercept coefficient of K4;
[0097] The sheep state perception data training set is divided into a training set, a validation set and a test set according to a certain ratio; the training set, the validation set and the test set are respectively input into a preset sheep state perception model for training, verification and testing; preferably, the sheep state perception data training set is scaled and normalized before being input into the maximum interval classification model to ensure that all the input data are of the same scale;
[0098] In an optional embodiment, the method further includes training a preset sheep state perception model:
[0099] The mutton state perception data set is a data set composed of several groups of mutton state perception data vectors; each mutton state perception data vector represents a sample point in a high-dimensional space, and each mutton state perception data vector corresponds to a label; in order to train and evaluate the performance of the model, the mutton state perception data training set is usually divided into three parts: a training set, a validation set and a test set; when dividing the data set, the first part (e.g. 80%) of the data set is generally used for model training, the second part (e.g. 10%) is used for model validation, and the remaining part (e.g. 10%) is used for model performance testing, and this division method can effectively evaluate the generalization ability and accuracy of the model;
[0100] Based on the maximum margin classification model, by selecting a suitable mapping function (such as K1, K2, K3 and K4, etc.) and combining the correction parameter C n , construct a preset sheep state perception model; the correction parameter C n The introduction of is used to control the complexity of the sheep state perception model, prevent overfitting and optimize the performance of the model; the correction parameter C n Control the degree of correction of the model for misclassified samples; correction parameter Cn The larger the correction parameter C is, the more serious the model's correction of misclassified samples is, and the higher the model complexity is; conversely, the correction parameter C n The smaller it is, the lighter the model's correction of misclassified samples is and the lower the model complexity is;
[0101] After the sheep state perception model training and verification are completed, the operating characteristic curve and confusion matrix are calculated using the test set data to evaluate the classification performance and accuracy of the model; according to the feedback of the operating characteristic curve and confusion matrix, the model is dynamically updated, and the model performance is improved by manually adjusting the model parameters or improving the model structure; the model is updated to obtain a new decision boundary; in the updating process, the correction parameter C is combined n Adjust the sheep state perception data vector to prevent overfitting; repeat this process until the preset stop condition is met, that is, the model performance reaches a certain threshold; through continuous training and updating, the sheep state perception model can be dynamically optimized to improve the perception ability and accuracy of the sheep state; according to actual needs, the sheep state perception model is applied to actual scenarios to classify and predict new unknown data;
[0102] The operating characteristic curve is a visualization tool for the performance of a classification model. Its horizontal axis represents different thresholds, and its vertical axis represents the relationship between the true positive rate and the false positive rate of the model. The true positive rate is the ratio of the number of samples correctly classified as the positive class to the number of samples in the actual positive class, while the false positive rate is the ratio of the number of samples incorrectly classified as the positive class to the number of samples in the actual negative class. The shape and position of the operating characteristic curve can help evaluate the performance of the model at different thresholds. The confusion matrix is a multidimensional matrix used to represent the prediction of different states and demonstrate the performance of the classification model. The confusion matrix presents the relationship between the actual category and the model predicted category in the matrix, and can be used to calculate various performance indicators, such as accuracy, precision, recall rate, and F1 score. The rows of the confusion matrix correspond to the actual categories, the columns correspond to the categories predicted by the model, and the value in each cell represents the number of samples that belong to the category of the row and are predicted to be in that category. By analyzing the confusion matrix, a more comprehensive understanding of the performance of the model on different categories can be obtained.
[0103] After testing on the dataset, the overall accuracy of the model is 91.92%, the overall precision of the model is 92.06%, the overall recall of the model is 92.14% and the overall F1 score of the model is 92.10%.
[0104] S205, based on obtaining new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment and constructing a new mutton sheep state perception data set, and inputting the new mutton sheep state perception data set into the constructed mutton sheep state perception model to obtain the state of the target mutton sheep in the feeding experiment;
[0105] The biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data in the new mutton sheep state perception data set are normalized, and the weight coefficients of different data are determined according to the contribution of the data; after the normalization processing of the new mutton sheep state perception data set, the weight input matrix is constructed in combination with the weight coefficients, and input into the mutton sheep state perception model;
[0106] Preferably, the sheep status perception process in the feeding experiment is as follows:
[0107] The weight input matrix enters the model through the input space; the weight input matrix is mapped to the high-dimensional space through the mapping space; in the mapping space, the decision boundary is defined to maximize the distance between data of different state categories; after mapping the state category results, the output is performed through the output space to obtain the output results; the core of the mutton sheep perception model is to accurately judge the state of the mutton sheep through the input raw data, so that the breeding personnel can timely discover abnormal conditions or confirm their normal conditions; preferably, in practical applications, the model can improve accuracy and reliability through continuous data input and model updating, so as to better cope with changes in the state of mutton sheep under different conditions;
[0108] The normalized calculation formula is as follows:
[0109]
[0110] Where: represents the normalized sheep state perception data vector; (i=1,2,3,…,m;k=bio,org,phy,env;j=1,2,3,4) represents the i-th row data vector in k parameter data;
[0111] The weight coefficient calculation formula is as follows:
[0112]
[0113] Where: Indicates P kThe weight coefficient of the parameter matrix (k = bio, org, phy, env); k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data respectively;
[0114] The weight input matrix calculation formula is as follows:
[0115]
[0116] Where: P bio , P org , P phy , P env They respectively represent the sheep state perception data matrix composed of biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data;
[0117] The output category formula is as follows:
[0118]
[0119] Where: R i Represents the output result of category i; A, B, C, D, E represent normal state, disease state, stress state, panic state and death state respectively; represents the decision space of category i consisting of the decision boundary;
[0120] Preferably, if the sheep is identified as being in an abnormal state during the feeding experiment, the knowledge base is called and personalized decisions are made in combination with the rich experience of the experimenters; the personalized decision includes determining the cause of the abnormal state, whether to continue the experiment, and whether to regulate the environment;
[0121] Based on the method in steps S101-S105, that is, the biochemical parameter data, vital sign parameter data and biological appearance parameter data of the experimental individuals in the feeding experiment and the environmental parameter data under the feeding experimental conditions are collected through the multi-scale sensor integrated unit technology, and after data processing, they are input into the mutton sheep state perception model, and the mutton sheep state results are further output; to represent the mutton sheep state prediction process, the overall steps are defined as follows:
[0122]
[0123] In the formula,
[0124]
[0125] They represent the biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data acquired from the start time t0 to the end time t respectively; t0 represents the start time of data acquisition; t represents the end time of data acquisition; P bio , P org , P phy , P env They respectively represent the sheep state perception data matrix composed of biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data; represents the normalized sheep state perception data vector; Indicates P k The weight coefficient of the parameter matrix (k = bio, org, phy, env); k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data respectively; W input represents the input matrix; MICM represents the maximum margin classification model; R i Represents the output result of category i;
[0126] Corresponding to the above method, such as Figure 4 As shown, this embodiment also provides a sheep state perception system in a feeding experiment, including:
[0127] The data acquisition module 401 is used to acquire basic information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and input the basic information into a preset knowledge base;
[0128] The perception data construction module 402 is used to extract data features from the vital sign parameter data respectively to obtain vital sign parameter feature data, and to construct a sheep state perception data set according to the biological appearance parameter data, vital sign parameter feature data, biochemical parameter data and environmental parameter data;
[0129] The training data construction module 403 is used to determine and classify the sheep state perception data set according to the threshold determination function, and to define labels for the data in the sheep state perception data set according to the determination and classification result, so as to construct a sheep state perception data training set;
[0130] The model training module 404 is used to input the mutton sheep state perception data training set into the preset mutton sheep state perception model for training, verification and testing to obtain a constructed mutton sheep state perception model;
[0131] The mutton sheep state perception module 405 is used to input the mutton sheep state perception data set into the constructed mutton sheep state perception model to obtain the state of the target mutton sheep in the feeding experiment.
[0132] The beneficial effects of the present invention are as follows:
[0133] (1) The present invention can accurately identify the behavioral characteristics of mutton sheep during breeding;
[0134] (2) The present invention can improve the welfare of mutton sheep during the mutton sheep breeding stage and the quality of mutton sheep breeding;
[0135] (3) The present invention can ensure the economic benefits of mutton breeding farms.
[0136] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0137] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A method for sensing the state of mutton sheep in a feeding experiment, characterized in that: include: Obtaining basic information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and inputting the basic information into a preset knowledge base; Performing window processing and data feature extraction on the vital sign parameter data to obtain vital sign parameter feature data, and constructing a sheep state perception data set based on the vital sign parameter feature data, biochemical parameter data, biological appearance parameter data and environmental parameter data; Determine and classify the sheep state perception data set according to a threshold determination function, and define labels for data in the sheep state perception data set according to the determination and classification result to construct a sheep state perception data training set; Input the mutton sheep state perception data training set into the preset mutton sheep state perception model for training, verification and testing to obtain a constructed mutton sheep state perception model; Based on obtaining new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target sheep in the feeding experiment, a new sheep state perception data set is constructed, and the new sheep state perception data set is input into the constructed sheep state perception model to obtain the state of the target sheep in the feeding experiment.
2. The method for sensing the state of mutton sheep in feeding experiments according to claim 1, characterized in that: The basic data information includes the number, strain, gender, age, health status and previous treatment of the target mutton sheep; the previous treatment refers to the treatment of the target mutton sheep before the feeding experiment begins.
3. The method for sensing the state of mutton sheep in feeding experiments according to claim 1, characterized in that: The biochemical parameter data include blood oxygen saturation, blood glucose concentration and blood lactate concentration; the vital sign parameter data include sound frequency change data, body temperature change data, heart rate change data and respiratory rate change data; the biological apparent parameter data include body weight, food intake and excretion; the environmental parameter data include noise, temperature, humidity, carbon dioxide concentration, oxygen concentration and carbon monoxide concentration.
4. The method for sensing the state of mutton sheep in feeding experiments according to claim 1, characterized in that: The vital sign parameter data is subjected to window processing and data feature extraction to obtain vital sign parameter feature data, and a sheep state perception data set is constructed according to the vital sign parameter feature data, biochemical parameter data, biological appearance parameter data and environmental parameter data, including: Extracting data features from the vital sign parameter data according to a preset vital sign parameter data feature extraction algorithm to obtain vital sign parameter feature data; The sheep state perception data matrix is constructed according to the biochemical parameter data, the biological appearance parameter data, the vital sign parameter characteristic data and the environmental parameter data, and each of the sheep state perception data matrices is assembled to obtain the sheep state perception data set.
5. The method for sensing the state of mutton sheep in feeding experiments according to claim 1, characterized in that: The feature calculation formula in the data feature extraction is: Where: x i represents the i-th data point in any windowed vital sign parameter data; N represents the total number of windowed vital sign parameter data; P(k) represents the power spectral density of the frequency spectrum of any windowed vital sign parameter data; N represents the total number of windowed vital sign parameter data; f(n) represents the amplitude of any windowed vital sign parameter data; x(n) represents the frequency of any windowed vital sign parameter data; p represents the exponential parameter; X(k) represents the complex amplitude spectrum of any windowed vital sign parameter data in the frequency domain; μ4 represents the fourth-order central moment of the frequency spectrum of any windowed vital sign parameter data; μ2 represents the second-order central moment of the frequency spectrum of any windowed vital sign parameter data.
6. The method for sensing the state of mutton sheep in feeding experiment according to claim 1, characterized in that: The expression of the sheep state perception data matrix includes: Where P bio , P org , P phy , P env The sheep state perception data matrix respectively represents the parameter matrix composed of the biochemical parameter data, the biological appearance parameter data, the vital sign parameter characteristic data and the environmental parameter data; represents the i-th row data vector in the k-parameter data, m means The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; T represents matrix transpose.
7. The method for sensing the state of mutton sheep in feeding experiment according to claim 1, characterized in that: The formula of the threshold determination function includes: In the formula, is the preset threshold; is the data vector of row i in the k parameter data matrix of the sheep state perception data, m means The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; represents the threshold range parameter; express The weight j parameter of express The weight l parameter; express The data in the meat sheep state perception data matrix with c rows and q columns; when In When it is within the threshold range, the data label is classified as normal; when In If it is not within the threshold range, the data label is classified as an abnormal state.
8. The method for sensing the state of mutton sheep in feeding experiment according to claim 1, characterized in that: The sheep state perception data training set is input into the preset sheep state perception model for training, verification and testing to obtain a constructed sheep state perception model, including: Dividing the sheep state perception data training set into a training set, a validation set and a test set according to a preset ratio; The training set, validation set and test set are respectively input into the preset mutton state perception model for training, validation and testing, and the optimal decision boundary is determined to obtain the constructed mutton state perception model; the preset mutton state perception model is constructed based on the maximum interval classification model; the maximum interval classification model is composed of input space, mapping space, decision boundary and output space; the maximum interval classification model is used to map the mutton state perception data training set to different category labels; the input space is the space of original data, and each feature vector is a sample as a point in the input space; the mapping space is constructed based on a mapping function; the mapping space is used to map the original data in the input space to a high-dimensional space to convert the original indivisible data into separable data; the decision boundary is a transcendental plane or transcendental surface that divides different categories in the mapping space to maximize the distance interval between different categories; the output space outputs the results classified by the decision boundary; the formula of the mapping function includes: Where: K p represents the mapping function; is the data vector of row i in the k parameter data matrix of the sheep state perception data, m means The total number of biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological apparent parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; ω v represents the weight coefficient of the mapping subfunction, (v=1,2,3,4); T represents the matrix transpose; θ and δ represent the slope coefficient of the mapping sub-function respectively; ε represents the intercept coefficient of the mapping sub-function; r represents the intercept coefficient of the mapping sub-function; the formula of the decision boundary definition function includes: Where: represents the decision boundary definition function; b * represents the bias coefficient of the decision boundary definition function; N represents the number of sheep state perception data vectors; Represents the sheep state perception data vector The category label of u Represents the sheep state perception data vector Lagrange multiplier; C n represents a correction parameter, which is used to control the complexity of the sheep state perception model; the correction parameter C n The update iterative process is as follows: n =C n-1 ·m n +c n ; Where: m n 、c n represents the randomly generated update parameter; C n Indicates the previous step value C n-1 Multiply by the update factor m n Add the update coefficient c n The result; the formula of the maximum interval definition function includes: Where: F Mar represents the maximum interval definition function; b represents the bias coefficient of the maximum interval definition function.
9. The method for sensing the state of mutton sheep in feeding experiment according to claim 1, characterized in that: Based on acquiring new biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment and constructing a new mutton sheep state perception data set, and inputting the new mutton sheep state perception data set into the constructed mutton sheep state perception model, the state of the target mutton sheep in the feeding experiment is obtained, including: The biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data in the new mutton sheep state perception data set are normalized, and the weight coefficients of different data are determined according to the contribution of the data; the calculation formula of the normalization processing includes: Where: represents the normalized sheep state perception data vector; is the data vector of row i in the k parameter data matrix of the sheep state perception data, m means The total number of biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data; k = bio, org, phy, env; bio, org, phy and env represent biochemical parameter data, biological appearance parameter data, vital sign parameter characteristic data and environmental parameter data respectively; j = 1, 2, 3, 4; N represents the number of mutton state perception data vectors; the weight coefficient calculation formula includes: Where: Indicates P k The weight coefficient of the parameter matrix; the weight input matrix calculation formula includes: Where: W input represents the weight matrix; P bio , P org , P phy , P env The sheep state perception data matrix respectively represents the parameter matrix composed of the biochemical parameter data, the parameter matrix composed of the biological appearance parameter data, the parameter matrix composed of the vital sign parameter characteristic data and the parameter matrix composed of the environmental parameter data; After the new mutton sheep state perception data set is normalized, a weight input matrix is constructed in combination with a weight coefficient and input into a mutton sheep state perception model; the formula of the output category of the mutton sheep state perception model includes: Where: R i represents the output result of category i; A, B, C, D, E respectively represent the normal state, disease state, stress state, panic state and death state of the target mutton sheep; represents the decision space of category i consisting of the decision boundary; If the meat sheep are identified as being in an abnormal state during the breeding experiment, the knowledge base will be called up and personalized decisions will be made based on the rich experience of the experimenters.
10. A sheep status perception system in a feeding experiment, characterized in that: include: A data acquisition module is used to acquire basic information of target mutton sheep in the feeding experiment and real-time and continuous biochemical parameter data, vital sign parameter data, biological appearance parameter data and environmental parameter data of the target mutton sheep in the feeding experiment, and input the basic information into a preset knowledge base; A perception data construction module is used to extract data features from the vital sign parameter data to obtain vital sign parameter characteristic data, and to construct a sheep state perception data set based on the biological appearance parameter data, the vital sign parameter characteristic data, the biochemical parameter data and the environmental parameter data; A training data construction module is used to determine and classify the sheep state perception data set according to a threshold determination function, and to define labels for the data in the sheep state perception data set according to the determination and classification result, so as to construct a sheep state perception data training set; A model training module is used to input the sheep state perception data training set into a preset sheep state perception model for training, verification and testing to obtain a constructed sheep state perception model; The mutton sheep state perception module is used to input the mutton sheep state perception data set into the constructed mutton sheep state perception model to obtain the state of the target mutton sheep in the feeding experiment.
Citation Information
Patent Citations
Mutton sheep feeding decision-making method and system based on dynamic monitoring
CN114493276A
Sheep reproductive health early warning system and method
CN116071783A
Technology for reducing methane yield in rumen of mutton sheep and daily ration preparation method thereof
CN116649483A
Remote monitoring system and method for beef cattle breeding
CN117379016A
Sheep behavior characteristic identification method and system
CN118296552A
Cited By
Mutton sheep barn feeding environment comfort level evaluation method based on physiological response
CN121144895A
