Ewe estrus stage identification method and system

Abstract

The invention provides an ewe estrus stage identification method and system, and relates to the technical field of precise and intelligent breeding, and the method comprises the steps: based on the ewe estrus stage identification method, obtaining genital tract physiological parameters and multi-dimensional motion data of ewes in a target breeding region, processing the genital tract physiological parameters and multi-dimensional motion data, constructing an estrus stage feature matrix, and carrying out estrus stage identification on the ewes in the target breeding region; and constructing a preset ewe estrus stage identification model based on a bionic element network algorithm, inputting the estrus stage feature matrix into the preset ewe estrus stage identification model, performing training, verification and testing, adjusting weight parameters according to a verification result, performing model evaluation according to a test result, and finally obtaining the ewe estrus stage identification model. And continuously acquiring new genital tract physiological parameters and multi-dimensional motion data in real time, constructing a new estrus stage feature matrix, and inputting the new estrus stage feature matrix into the ewe estrus stage identification model to obtain the estrus stage of the ewes in the target breeding area. The estrus stage of the ewe can be identified timely and accurately.

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 identifying the estrus stage of ewes. Background Art

[0002] Modern animal husbandry is developing in the direction of scale and intensification. Timely and accurate identification of the estrus stage of ewes can ensure the best breeding time, increase the birth rate of healthy lambs, and thus improve the economic benefits of the farm. By effectively monitoring and identifying the estrus stage of ewes, farms can reduce unnecessary waste of resources, reduce breeding costs, and thus improve the benefits of the entire breeding process. With the development of science and technology, the introduction of advanced sensing technology and neural network algorithms has significant significance and advantages for estrus monitoring and identification of ewes. Advanced sensors can monitor the physiological parameters of ewes in real time and accurately obtain data related to different estrus stages. Deep learning models can learn features from massive data and realize accurate estrus stage identification. Compared with traditional manual monitoring and empirical judgment, the application of this comprehensive new technology can greatly improve the accuracy and effectiveness of estrus stage monitoring and identification.

[0003] Therefore, there is an urgent need for a method and system for identifying the estrus stage of ewes. By integrating data from multiple sensors and advanced deep learning technology, it can timely and accurately identify the estrus stage of ewes and provide scientific basis and intelligent decision support. This method can not only improve breeding efficiency and economic benefits, but also help promote the development of the breeding industry towards digitalization and intelligence, thereby meeting the needs of modern breeding industry for efficient management and production. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a method and system for identifying the estrus stage of a ewe.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] The present invention provides a method for identifying the estrus stage of a ewe, comprising:

[0007] Obtain reproductive tract physiological parameters and multi-dimensional movement data of ewes in the target breeding area;

[0008] Processing the reproductive tract physiological parameters and multi-dimensional motion data to construct an estrus stage characteristic matrix;

[0009] Construct a preset ewe estrus stage recognition model based on the bionic meta-network algorithm;

[0010] Input the estrus stage feature matrix into the preset ewe estrus stage identification model for training, verification and testing; adjust the weight parameters according to the verification results, evaluate the model according to the test results, and finally obtain the ewe estrus stage identification model;

[0011] New reproductive tract physiological parameters and multi-dimensional motion data are continuously acquired in real time and a new estrus stage characteristic matrix is ​​constructed; the new estrus stage characteristic matrix is ​​input into the estrus stage recognition model of the ewe to obtain the estrus stage of the ewes in the target breeding area.

[0012] Preferably, the reproductive tract physiological parameters include reproductive tract mucosal impedance, serum estradiol concentration, progesterone, reproductive tract temperature, reproductive tract humidity, reproductive tract pH value, serum follicle-stimulating hormone, and reproductive tract secretion viscosity;

[0013] The multi-dimensional motion data includes acceleration motion data, angular velocity motion data and angle motion data.

[0014] Preferably, the reproductive tract physiological parameters and multi-dimensional motion data are processed to construct an estrus stage characteristic matrix, specifically:

[0015] The missing values, abnormal values ​​and noise values ​​in the reproductive tract physiological parameters are processed. After the processing is completed, the time series feature extraction is performed for each reproductive tract physiological parameter to construct a reproductive tract physiological parameter feature matrix, wherein the time series feature extraction formula is:

[0016]

[0017]

[0018] In the formula, A1, A2, A3, A4 are time series features, x i is the ith data of reproductive tract physiological parameters, x max is the largest data point among reproductive tract physiological parameters, x min is the minimum data point in reproductive tract physiological parameters, and N is the total number of reproductive tract physiological parameter data;

[0019] The multi-dimensional motion data is subjected to window processing to construct a multi-dimensional motion window data set, and motion features are extracted to construct a multi-dimensional motion feature matrix, wherein the motion feature extraction formula is:

[0020]

[0021] In the formula, B1, B2, B3, and B4 are motion characteristics. x(t) is the original data, f is the frequency, j is an imaginary number, N is the number of samples of the data, T is the total length of the data, Δf is the frequency interval, k1 and k2 are the starting indexes of the frequency range;

[0022] Based on the reproductive tract physiological parameter feature matrix and the multi-dimensional motion feature matrix, the estrus stage feature matrix is ​​constructed, and the labels of the estrus stage feature matrix are defined to construct the estrus stage feature input matrix, in which the labels are different estrus stages, including proestrus, estrus, metestrus and diestrus.

[0023] Preferably, the structure of the preset ewe estrus stage identification model is an input layer, a hidden layer and an output layer in sequence, wherein the input layer is composed of a plurality of bionic elements and is responsible for receiving external data and transmitting it to the network, the hidden layer is located between the input layer and the output layer and is responsible for processing the input data, and the output layer is responsible for generating the final output result;

[0024] Among them, the bionic element formula of the input layer is:

[0025]

[0026] In the formula, a i is the i-th bionic element vector, X i is the input data sequence vector, W Q , W K , W V are the learned weight matrices, which are query weight matrix, key weight matrix and value weight matrix, respectively. k is the dimension of the kth key vector, n is the total number, T is the matrix transpose operation, W o is the output weight matrix, which is used to map the connected multi-head results to the output space of the input layer;

[0027] The bionic element formula of the hidden layer is:

[0028]

[0029] In the formula, is the output of the lth layer based on the ith bionic element, is the weight matrix from the hidden layer to the output layer, is the bias term from the hidden layer to the output layer, is the positive exponent of the activation value obtained by weighted input, The negative exponential of the activations obtained by weighting the input creates symmetry, is the weight matrix input to the hidden layer in the lth layer, is the weight matrix from the previous hidden state to the current hidden state in the lth layer, is the hidden state of the lth layer of the i-th bionic neural unit at time step (t-1), is the bias term of the lth layer;

[0030] The output formula of the output layer is:

[0031]

[0032] In the formula, is the output of the i-th bionic element in the l-th layer, is the weight coefficient from the qth bionic neural unit in the l-1th layer to the ith bionic element in the lth layer, is the output of the l-1th layer of the hidden layer based on the i-th bionic element.

[0033] Preferably, the estrus stage feature matrix is ​​input into a preset ewe estrus stage recognition model for training, verification and testing; the weight parameters are adjusted according to the verification results, and the model is evaluated according to the test results, and finally the ewe estrus stage recognition model is obtained, specifically:

[0034] The estrus stage feature matrix is ​​divided into a training matrix, a verification matrix and a test matrix in proportion, and the training matrix is ​​input into the preset ewe estrus stage recognition model for training. Then, after obtaining the network weights, the verification matrix is ​​used to verify the trained preset ewe estrus stage recognition model, and the weight parameters are adjusted according to the verification results. The verified preset ewe estrus stage recognition model is tested and evaluated according to the test matrix, and finally the ewe estrus stage recognition model is obtained.

[0035] The present invention also provides a system for identifying estrus stages of ewes, comprising:

[0036] Data acquisition module: used to collect real-time physiological parameters and multi-dimensional movement data of the ewe's reproductive tract;

[0037] Data processing module: used to process the collected data;

[0038] Estrus stage identification module: used to identify the estrus stage of ewes.

[0039] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0040] The present invention provides a method and system for identifying the estrus stage of ewes. The method comprises obtaining reproductive tract physiological parameters and multidimensional motion data of ewes in a target breeding area, processing the reproductive tract physiological parameters and multidimensional motion data, constructing an estrus stage characteristic matrix, constructing a preset ewe estrus stage identification model based on a bionic meta-network algorithm, inputting the estrus stage characteristic matrix into the preset ewe estrus stage identification model, and performing training, verification and testing; adjusting weight parameters according to the verification results, performing model evaluation according to the test results, and finally obtaining the ewe estrus stage identification model, obtaining new reproductive tract physiological parameters and multidimensional motion data in real time and continuously and constructing a new estrus stage characteristic matrix; inputting the new estrus stage characteristic matrix into the ewe estrus stage identification model, and obtaining the estrus stage of ewes in the target breeding area. The present invention realizes real-time and accurate monitoring and identification of the estrus stage of ewes; by comprehensively utilizing reproductive tract physiological parameters and multi-dimensional motion data, combined with advanced data processing and analysis technology, the present invention can effectively identify different estrus stages of ewes, thereby providing a scientific basis for breeding management; the system greatly improves the accuracy and efficiency of estrus monitoring through automated data collection, processing and analysis, and avoids the influence of manual intervention and observation errors in traditional methods; in addition, the system of the present invention also has strong real-time performance, can continuously track the physiological and behavioral changes of ewes, timely discover key changes in the estrus cycle, and help breeders seize the best breeding time; by accurately identifying the estrus stage, breeding decisions are optimized, reproductive efficiency is effectively improved, resource waste within the breeding cycle is reduced, breeding costs are reduced, and production benefits are improved; in general, the estrus stage identification method and system of the present invention not only improves the scientificity and intelligent level of breeding management, but also provides a more efficient and sustainable solution for modern breeding industry, which is of great significance to promoting the digitalization and intelligent development of agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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.

[0042] Figure 1 A schematic flow chart of a method for identifying the estrus stage of a ewe provided in an embodiment of the present invention;

[0043] Figure 2 A detailed schematic diagram of a method for identifying the estrus stage of a ewe provided in an embodiment of the present invention;

[0044] Figure 3A schematic diagram of the structure of a system for identifying the estrus stage of ewes provided in an embodiment of the present invention.

[0045] Reference numerals: 201, data acquisition module; 202, data processing module; 203, estrus stage identification module. DETAILED DESCRIPTION

[0046] 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.

[0047] The purpose of the present invention is to provide a method and system for identifying the estrus stage of a ewe, which can timely and accurately identify the estrus stage of a ewe by integrating data from multiple sensors and advanced deep learning technology.

[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, Figure 2 A detailed flow chart of the method for identifying the estrus stage of ewes provided by an embodiment of the present invention is as follows: Figure 1 and Figure 2 As shown, the present invention provides a method for identifying the estrus stage of a ewe, comprising:

[0050] Step 100: Acquire reproductive tract physiological parameters and multi-dimensional movement data of ewes in the target breeding area;

[0051] Step 200: Processing the reproductive tract physiological parameters and multi-dimensional motion data to construct an estrus stage characteristic matrix;

[0052] Step 300: constructing a preset ewe estrus stage recognition model based on a bionic meta-network algorithm;

[0053] Step 400: input the estrus stage feature matrix into a preset ewe estrus stage identification model for training, verification and testing; adjust the weight parameters according to the verification results, perform model evaluation according to the test results, and finally obtain the ewe estrus stage identification model;

[0054] Step 500: Real-time and continuous acquisition of new reproductive tract physiological parameters and multi-dimensional motion data and construction of a new estrus stage characteristic matrix; inputting the new estrus stage characteristic matrix into the ewe estrus stage recognition model to obtain the estrus stage of the ewes in the target breeding area.

[0055] In step 100, the reproductive tract physiological parameters include but are not limited to reproductive tract mucosal impedance, serum estradiol concentration, progesterone, reproductive tract temperature, reproductive tract humidity, reproductive tract pH value, serum follicle-stimulating hormone, and reproductive tract secretion viscosity;

[0056] The multi-dimensional motion data includes but is not limited to acceleration motion data, angular velocity motion data and angle motion data.

[0057] It should be noted that the present invention provides an embodiment, which preferably obtains the reproductive tract physiological parameters and the multi-dimensional motion data of the ewe at different estrus stages;

[0058] For the convenience of classification, the present invention divides the estrus stage of ewes into proestrus, estrus, metestrus and diestrus. It should be noted that these four estrus stages do not cover all estrus stages of ewes, but are simplified classifications for the convenience of the subsequent explanation of specific implementation plans; in practical applications, the estrus cycle of ewes may be more finely divided into multiple different stages. Considering the actual situation, the implementation of the present invention is not limited to the division of the above four stages, and other subdivided estrus stages can also be included in the technical scope of the present invention; therefore, in the subsequent implementation process, it can be flexibly adjusted and expanded according to actual needs and physiological characteristics of the research object;

[0059] The specific definitions of the four estrus stages of ewes are shown in Table 1.

[0060] Table 1 Specific definitions of the four estrus stages of ewes

[0061]

[0062]

[0063] It should be noted that in order to ensure the authenticity and reliability of the measured data and the robustness of the multi-identification model, the basic biological characteristics, breeding environment, feeding management and nutritional level of the monitored sheep must be consistent; the basic biological characteristics include but are not limited to the type and age of the sheep, to ensure that the physiological state and characteristics of each sheep are within a comparable range.

[0064] In step 200, the reproductive tract physiological parameters and multi-dimensional motion data are processed to construct an estrus stage characteristic matrix, specifically:

[0065] The missing values, abnormal values ​​and noise values ​​in the reproductive tract physiological parameters are processed to ensure the quality of the data. After the processing is completed, the time series feature extraction is performed for each reproductive tract physiological parameter to construct the reproductive tract physiological parameter feature matrix, where the time series feature extraction formula is:

[0066]

[0067] In the formula, A1, A2, A3, A4 are time series features, x i is the ith data of reproductive tract physiological parameters, x max is the largest data point among reproductive tract physiological parameters, x max is the minimum data point in reproductive tract physiological parameters, and N is the total number of reproductive tract physiological parameter data;

[0068] The multi-dimensional motion data is subjected to window processing to construct a multi-dimensional motion window data set, and motion features are extracted to construct a multi-dimensional motion feature matrix, wherein the motion feature extraction formula is:

[0069]

[0070]

[0071] In the formula, B1, B2, B3, and B4 are motion characteristics. x(t) is the original data, f is the frequency, j is an imaginary number, N is the number of samples of the data, T is the total length of the data, Δf is the frequency interval, k1 and k2 are the starting indexes of the frequency range;

[0072] Based on the reproductive tract physiological parameter feature matrix and the multi-dimensional motion feature matrix, the estrus stage feature matrix is ​​constructed, and the labels of the estrus stage feature matrix are defined to construct the estrus stage feature input matrix, in which the labels are different estrus stages, including proestrus, estrus, metestrus and diestrus.

[0073] In step 300, the structure of the preset ewe estrus stage identification model is sequentially an input layer, a hidden layer and an output layer. The input layer is composed of a plurality of bionic elements, which is mainly responsible for receiving external input data (such as an estrus stage feature matrix) and passing these data to the network for further processing. The design of the input layer is to convert the original data into a format suitable for bionic element network processing to ensure the accuracy and effectiveness of data transmission; the hidden layer is located between the input layer and the output layer, and its main function is to perform multi-level nonlinear transformation and feature extraction on the input data. Through the hidden layer, the model can automatically learn the important patterns and laws in the input data so as to better identify the estrus stage. The structure of the hidden layer is usually composed of a plurality of bionic elements and an activation function. These neural units transmit information through weighted connections, and introduce nonlinear characteristics through the activation function, thereby enhancing the expression ability of the model; the output layer is responsible for generating the final output result, which usually corresponds to the identification result of the current estrus stage of the ewe; in the output layer, the bionic element network summarizes the processed data and outputs a classification result representing the estrus stage and its probability value, thereby completing the identification of the estrus state of the ewe;

[0074] Among them, the bionic element formula of the input layer is:

[0075]

[0076] In the formula, a i is the i-th bionic element vector, X i is the input data sequence vector, W Q , W K , W V are the learned weight matrices, which are query weight matrix, key weight matrix and value weight matrix, respectively. k is the dimension of the kth key vector, n is the total number, T is the matrix transpose operation, W o is the output weight matrix, which is used to map the connected multi-head results to the output space of the input layer;

[0077] The bionic element formula of the hidden layer is:

[0078]

[0079] In the formula, is the output of the lth layer based on the i-th bionic neural unit, is the weight matrix from the hidden layer to the output layer, is the bias term from the hidden layer to the output layer, is the positive exponent of the activation value obtained by weighted input, The negative exponential of the activations obtained by weighting the input creates symmetry, is the weight matrix input to the hidden layer in the lth layer, is the weight matrix from the previous hidden state to the current hidden state in the lth layer, is the hidden state of the lth layer of the i-th bionic neural unit at time step (t-1), is the bias term of the lth layer;

[0080] The output formula of the output layer is:

[0081]

[0082] In the formula, is the output of the i-th bionic element in the l-th layer, is the weight coefficient from the qth bionic neural unit in the l-1th layer to the ith bionic element in the lth layer, is the output of the l-1th layer of the hidden layer based on the i-th bionic element.

[0083] In step 400, the estrus stage feature matrix is ​​input into a preset ewe estrus stage identification model for training, verification and testing; the weight parameters are adjusted according to the verification results, and the model is evaluated according to the test results, and finally the ewe estrus stage identification model is obtained, which is specifically:

[0084] The estrus stage feature matrix is ​​divided into training matrix, validation matrix and test matrix in proportion to ensure the reasonable distribution of the data set and the accuracy of model evaluation. In the data preprocessing stage, the training matrix is ​​used to input into the preset ewe estrus stage recognition model as training data to optimize the model parameters. Through training methods such as back propagation algorithm and gradient descent, the model adjusts the internal weights in each iteration to minimize the prediction error and gradually learns the best features to identify the estrus stage of ewes. After the training is completed, the network weights will be saved and the model will be verified using the validation matrix. The verification process can check the performance of the trained model on unknown data. Verification results: if the performance of the model does not meet expectations, the weight parameters may be adjusted or the network structure may be optimized according to the verification error to improve its recognition ability; after the model training and verification process is completed, the optimized ewe estrus stage recognition model is finally tested and evaluated using the test matrix; the data in the test matrix is ​​used to simulate the scenarios in real applications and evaluate the generalization ability of the model on unseen data; based on the test results, the accuracy, precision, recall rate and other indicators of the model are further evaluated to determine whether it meets the actual application requirements; finally, through these training, verification and testing steps, a fully optimized and verified ewe estrus stage recognition model is obtained;

[0085] The number of layers of the preset ewe estrus stage recognition model based on the bionic element network algorithm is set to l+1 (l≥2), and the number of bionic elements in each layer from the input layer to the output layer is m0, m1, m2, ..., m l+1 ; m0 is the dimension of the input estrus stage feature vector; m l is the number of estrus stages to be output; m1, m2, ..., m l is the number of bionic elements in each hidden layer; the output vectors of each layer of the preset ewe estrus stage identification model are expressed as follows:

[0086] Input Layer:

[0087] Hidden layer 1:

[0088] Hidden layer 2:

[0089] …;

[0090] Hidden layer l:

[0091] Output layer:

[0092] The weight matrix and bias term of each hidden layer are defined based on the output vector of each hidden layer;

[0093] Hidden layer 1:

[0094] Hidden layer 2:

[0095] …;

[0096] Hidden layer l:

[0097] The output error is calculated based on the concept of error back propagation; the weight matrix and bias term of each hidden layer are calculated and updated based on the output error; the output error calculation formula is as follows:

[0098]

[0099] In the formula, is the output error, is the difference between the ith output vector of the lth hidden layer and the ith input vector of the input layer, and ⊙ is the element-wise product;

[0100] The calculation formula for the reverse error to each hidden layer is as follows:

[0101]

[0102] In the formula, is the reverse error of the lth hidden layer, Represents the weight matrix from the lth hidden layer to the output layer

[0103] The update formulas for the weight matrix and bias term are as follows:

[0104]

[0105] In the formula, are the updated weight matrix and bias term respectively, is the learning factor, l is the number of hidden layers, A i The i-th input vector of the input layer, is the reverse error of the lth hidden layer;

[0106] By continuously adjusting the weight matrix and bias terms, and repeatedly updating the model parameters and performing cross-validation, the performance of the model can be effectively improved until the expected effect is achieved. In each iteration, the model will update the parameters through the training matrix, and use the cross-validation method to evaluate the performance of the model under different parameter settings, thereby minimizing overfitting and underfitting phenomena and ensuring the generalization ability and stability of the model. After multiple rounds of optimization and verification, the trained and verified model is finally tested and evaluated using the test matrix. In the test phase, a series of performance indicators, including accuracy, precision, recall rate and F1 score, can be obtained by calculating the confusion matrix of the model. These indicators will help evaluate the performance of the model in different estrus stage identification tasks; according to the results of the confusion matrix, especially the specific circumstances of misclassification, the parameters of the ewe estrus stage identification model can be actively adjusted, such as adjusting the weights, updating the learning rate or selecting different regularization strategies, to further improve the accuracy and robustness of the model; the core of this process is to obtain an optimal ewe estrus stage identification model through repeated model tuning and parameter adjustment.

[0107] The test matrix shows that the overall model accuracy is 91.22%, the overall model precision is 90.51%, the overall model recall is 91.44% and the overall model F1 score is 90.91%.

[0108] In step 500, new reproductive tract physiological parameters and multi-dimensional motion data are continuously acquired in real time and a new estrus stage characteristic matrix is ​​constructed; the new estrus stage characteristic matrix is ​​input into the estrus stage recognition model of the ewe to obtain the estrus stage of the ewes in the target breeding area, specifically:

[0109] The new estrus stage feature matrix is ​​input into the estrus stage recognition model of the ewe in real time. With the help of the powerful predictive ability of the model, the current estrus stage of each ewe is quickly and accurately identified. By real-time analysis of the latest collected physiological and movement data, the model can timely determine the estrus state of each ewe, thereby providing important information about the reproductive cycle of the ewe. This process not only ensures the real-time and continuity of the data flow, but also can dynamically adjust and optimize the recognition results at different time points. The real-time monitoring system provides scientific decision-making support for breeding managers, helping breeding personnel to grasp the estrus status of ewes in a timely manner, thereby effectively optimizing estrus monitoring and breeding decisions. Through accurate estrus stage identification, the best breeding time can be accurately grasped, the breeding success rate can be improved, and unnecessary human intervention can be reduced, thereby improving the efficiency and economic benefits of the entire breeding process. Ultimately, the application of this system will significantly improve the breeding management process of ewes, ensuring that it is carried out in a more scientific, efficient and sustainable manner.

[0110] like Figure 3 As shown, the present invention also provides a system for identifying estrus stages of ewes, comprising:

[0111] Data acquisition module 201: This module has multiple functions of data information collection, storage and transmission. By connecting with sensor equipment, it collects the reproductive tract physiological parameters and movement data of ewes in real time. The collected data will be stored in the local database to ensure the integrity and reliability of the data. At the same time, the module also supports the transmission of data to the data processing module or cloud platform by wireless or wired means for subsequent analysis and processing;

[0112] Data processing module 202: This module is mainly responsible for correcting and compensating the collected raw data information to eliminate the interference of noise, abnormal values ​​and missing values ​​to ensure the high quality of data. By adopting data cleaning and preprocessing technology, the module can effectively improve the accuracy of data and convert it into a format suitable for model analysis. In addition, the module also has the ability to process multidimensional data, including time series feature extraction, data standardization and feature selection, etc., and ultimately provide reliable data input for estrus stage identification;

[0113] Estrus stage identification module 203: This module is the core part of the system, responsible for receiving, storing, analyzing and determining the estrus stage of the ewe. The received data will be analyzed in real time through the trained model, combined with the bionic meta-network algorithm, to accurately identify the physiological state and behavioral characteristics of the ewe. The output of the module will be the current estrus stage of the ewe, and detailed status determination results can be provided to provide a scientific decision-making basis for breeding managers.

[0114] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0115] 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 identifying the estrus stage of a ewe, characterized in that: include: Obtain reproductive tract physiological parameters and multi-dimensional movement data of ewes in the target breeding area; Processing the reproductive tract physiological parameters and multi-dimensional motion data to construct an estrus stage characteristic matrix; Construct a preset ewe estrus stage recognition model based on the bionic meta-network algorithm; Input the estrus stage feature matrix into the preset ewe estrus stage identification model for training, verification and testing; adjust the weight parameters according to the verification results, evaluate the model according to the test results, and finally obtain the ewe estrus stage identification model; Acquire new reproductive tract physiological parameters and multi-dimensional motion data in real time and continuously and construct a new estrus stage characteristic matrix; The new estrus stage characteristic matrix is ​​input into the estrus stage recognition model of the ewe to obtain the estrus stage of the ewes in the target breeding area.

2. The method for identifying the estrus stage of a ewe according to claim 1, characterized in that: The reproductive tract physiological parameters include reproductive tract mucosal impedance, serum estradiol concentration, progesterone, reproductive tract temperature, reproductive tract humidity, reproductive tract pH value, serum follicle-stimulating hormone, and reproductive tract secretion viscosity; The multi-dimensional motion data includes acceleration motion data, angular velocity motion data and angle motion data.

3. The method for identifying the estrus stage of a ewe according to claim 1, characterized in that: The reproductive tract physiological parameters and multi-dimensional motion data are processed to construct an estrus stage characteristic matrix, specifically: The missing values, abnormal values ​​and noise values ​​in the reproductive tract physiological parameters are processed. After the processing is completed, the time series feature extraction is performed for each reproductive tract physiological parameter to construct a reproductive tract physiological parameter feature matrix, wherein the time series feature extraction formula is: In the formula, A1, A2, A3, A4 are time series features, x i is the ith data of reproductive tract physiological parameters, x max is the largest data point among reproductive tract physiological parameters, x min is the minimum data point in reproductive tract physiological parameters, and N is the total number of reproductive tract physiological parameter data; The multi-dimensional motion data is subjected to window processing to construct a multi-dimensional motion window data set, and motion features are extracted to construct a multi-dimensional motion feature matrix, wherein the motion feature extraction formula is: In the formula, B1, B2, B3, and B4 are motion characteristics. x(t) is the original data, f is the frequency, j is an imaginary number, N is the number of samples of the data, T is the total length of the data, Δf is the frequency interval, k1 and k2 are the starting indexes of the frequency range; Based on the reproductive tract physiological parameter feature matrix and the multi-dimensional motion feature matrix, the estrus stage feature matrix is ​​constructed, and the labels of the estrus stage feature matrix are defined to construct the estrus stage feature input matrix, in which the labels are different estrus stages, including proestrus, estrus, metestrus and diestrus.

4. The method for identifying the estrus stage of a ewe according to claim 1, characterized in that: The structure of the preset ewe estrus stage identification model is sequentially composed of an input layer, a hidden layer and an output layer, wherein the input layer is composed of a plurality of bionic elements and is responsible for receiving external data and transmitting it to the network, the hidden layer is located between the input layer and the output layer and is responsible for processing the input data, and the output layer is responsible for generating the final output result; Among them, the bionic element formula of the input layer is: In the formula, a i is the i-th bionic element vector, X i is the input data sequence vector, W Q , W K , W V are the learned weight matrices, which are query weight matrix, key weight matrix and value weight matrix, respectively. k is the dimension of the kth key vector, n is the total number, T is the matrix transpose operation, W o is the output weight matrix, which is used to map the connected multi-head results to the output space of the input layer; The bionic element formula of the hidden layer is: In the formula, is the output of the lth layer based on the ith bionic element, is the weight matrix from the hidden layer to the output layer, is the bias term from the hidden layer to the output layer, is the positive exponent of the activation value obtained by weighted input, The negative exponential of the activations obtained by weighting the input creates symmetry, is the weight matrix input to the hidden layer in the lth layer, is the weight matrix from the previous hidden state to the current hidden state in the lth layer, is the hidden state of the lth layer of the i-th bionic neural unit at time step (t-1), is the bias term of the lth layer; The output formula of the output layer is: In the formula, is the output of the i-th bionic element in the l-th layer, is the weight coefficient from the qth bionic neural unit in the l-1th layer to the ith bionic element in the lth layer, is the output of the l-1th layer of the hidden layer based on the i-th bionic element.

5. The method for identifying the estrus stage of a ewe according to claim 1, characterized in that: The estrus stage feature matrix is ​​input into the preset ewe estrus stage identification model for training, verification and testing; the weight parameters are adjusted according to the verification results, and the model is evaluated according to the test results, and finally the ewe estrus stage identification model is obtained, which is specifically: The estrus stage feature matrix is ​​divided into a training matrix, a verification matrix and a test matrix in proportion, and the training matrix is ​​input into the preset ewe estrus stage recognition model for training. Then, after obtaining the network weights, the verification matrix is ​​used to verify the trained preset ewe estrus stage recognition model, and the weight parameters are adjusted according to the verification results. The verified preset ewe estrus stage recognition model is tested and evaluated according to the test matrix, and finally the ewe estrus stage recognition model is obtained.

6. A system for identifying estrus stage of ewes, characterized in that: include: Data acquisition module: used to collect real-time physiological parameters and multi-dimensional movement data of the ewe's reproductive tract; Data processing module: used to process the collected data; Estrus stage identification module: used to identify the estrus stage of ewes.

Images