CN114417242A - Big data detection system for livestock and poultry activity information - Google Patents

Abstract

The invention discloses a big data detection system for livestock and poultry activity information, which consists of a cloud platform-based livestock and poultry sign parameter acquisition and intelligent prediction platform and a livestock and poultry activity big data prediction subsystem, wherein the cloud platform-based livestock and poultry sign parameter acquisition and intelligent prediction platform realizes the detection and processing of the temperature and activity information parameters of livestock and poultry signs; the livestock and poultry activity big data prediction subsystem realizes prediction of the activity state of livestock and poultry and provides data and early warning for preventing livestock and poultry diseases; the invention aims to provide a big data detection system for livestock and poultry activity information, which monitors the temperature and activity information of livestock and poultry bodies in real time, thereby providing data and early warning for preventing livestock and poultry diseases.

Description

Big data detection system for livestock and poultry activity information

technical field

The invention relates to the technical field of livestock and poultry activity information parameter detection, in particular to a big data detection system for livestock and poultry activity information.

Background technique

As an important part of my country's national economic industry, livestock and poultry breeding industry, how to meet people's increasingly diverse needs for livestock and poultry products, has put forward new requirements for the breeding process and breeding standards of the entire livestock and poultry breeding industry. In the traditional livestock and poultry breeding process, simple, rough, and inefficient high-frequency manual intervention breeding methods are often used, and it is often impossible to obtain information on the activities of breeding individuals in a timely manner after spending a lot of labor and time costs. The artificial participation of the farm has also caused a certain impact on the physiological activities of the breeding individuals to a certain extent, increasing the probability of cross-infection between humans and animals. In view of such drawbacks, it is more and more urgent to need modern breeding technology to promote the breeding industry. Further development, so that the livestock and poultry industry can withstand the new test of contemporary society. The invention develops a livestock and poultry miniature wearable activity sign sensing device, builds a data transmission network based on wireless network technology, develops platform software for monitoring data analysis and processing, and realizes the real-time display of livestock and poultry activity data. Livestock and poultry diseases provide data and early warning.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a big data detection system for livestock and poultry activity information, which can monitor the body temperature and activity information of livestock and poultry in real time, thereby providing data and early warning for the prevention of livestock and poultry diseases.

The present invention is achieved through the following technical solutions:

A big data detection system for livestock and poultry activity information, which realizes the detection of livestock and poultry body temperature and activity information parameters and the intelligent prediction of livestock and poultry posture, the big data detection system includes a cloud platform-based livestock and poultry sign parameter collection and intelligent prediction platform and Livestock Activity Big Data Prediction Subsystem.

The further technical improvement scheme of the present invention is:

The cloud platform-based livestock and poultry sign parameter collection and intelligent prediction platform is composed of detection nodes, gateway nodes, on-site monitoring terminals, cloud platforms and mobile APPs. The detection nodes collect the body temperature and activity information parameters of livestock and poultry and upload them to the cloud platform through the gateway nodes. Store data and publish information on the cloud platform. The mobile phone APP can monitor the signs and motion parameters of livestock and poultry in real time through the body temperature and activity information provided by the cloud platform. The detection node is responsible for collecting the body temperature and activity information parameters of livestock and poultry, which is realized through the gateway node The two-way communication between the detection node, on-site monitoring terminal, cloud platform and mobile APP realizes the collection and prediction of livestock and poultry body temperature and activity information parameters; the structure diagram of the livestock and poultry physical parameter collection and intelligent prediction platform based on the cloud platform is shown in Figure 1.

The further technical improvement scheme of the present invention is:

The livestock and poultry activity big data prediction subsystem is composed of 3 parameter detection modules, BAM neural network model, 2 beat delay line TDL, 2 ARIMA prediction models and ART2 neural network model parts. The three-axis acceleration sensor ADXL362 senses the detected livestock. The accelerations of the bird's X, Y and Z axes are respectively used as the input of the corresponding parameter detection module, the 3 connection numbers output by the 3 parameter detection modules are used as the input of the BAM neural network model, and the binary connection number output by the BAM neural network model. As the input corresponding to the ART2 neural network model, the determined value and the fluctuation value output by the ART2 neural network model are used as the input of the two beat delay lines TDL and the corresponding input of the ART2 neural network model, and the two beat delay line TDL outputs are respectively used as the corresponding input The input of the two ARIMA prediction models, the outputs of the two ARIMA prediction models are used as the corresponding inputs of the ART2 neural network model. The definite value c and the fluctuation value d output by the ART2 neural network model constitute a binary connection number c+di, and the ART2 neural network model The output binary connection number corresponds to 5 different states of the livestock and poultry being crawling, walking, standing, lying down and lying on the side respectively; the structure diagram of the big data prediction subsystem of livestock and poultry activity is shown in Figure 2.

The further technical improvement scheme of the present invention is:

The parameter detection module is composed of Adaline neural network model with time delay unit, EMD empirical mode decomposition model, GM (1, 1) gray prediction model, multiple NARX neural network prediction models, 2 beat delay lines TDL, 2 ARIMA The prediction model and the wavelet neural network model of the binary connection number are composed; the output of the parameter detection sensor is used as the input of the Adaline neural network model with time delay unit, and the output of the Adaline neural network model with time delay unit is used as the input of the EMD empirical mode decomposition model. , the low-frequency trend value of the measured parameter output by the EMD empirical modal decomposition model is used as the input of the GM(1,1) gray prediction model, and the different high-frequency fluctuation values of the measured parameter output by the EMD empirical modal decomposition model are used as the corresponding The input of multiple NARX neural network prediction models, the output of GM(1, 1) gray prediction model and the output of multiple NARX neural network prediction models are respectively used as the corresponding input of the wavelet neural network model of the binary connection number, and the wavelet of the binary connection number The neural network model outputs the definite value a and the fluctuation value b of the measured parameter to form the binary connection coefficient a+bi of the measured parameter size, and the definite value a and the fluctuation value b of the measured parameter are used as the The input and the two corresponding inputs of the wavelet neural network model of the binary connection number, the two TDL outputs of the beat delay line are respectively used as the input of the corresponding ARIMA prediction model, and the outputs of the two ARIMA prediction models are respectively used as the wavelet neural network of the binary connection number. The corresponding input of the network model, the binary connection number output by the wavelet neural network model of the binary connection number is used as the predicted value of the measured parameter. The structure diagram of the parameter detection module is shown in Figure 3.

Compared with the prior art, the present invention has the following obvious advantages:

1. The present invention decomposes the livestock and poultry activity information sequence output by the original Adaline neural network model with time-delay unit into components of different frequency bands through the EMD empirical mode decomposition model, and each component shows different features hidden in the original sequence. information. to reduce the non-stationarity of the series. The data of the high-frequency fluctuation part of the livestock and poultry activity process are not closely related, and the frequency is relatively high, which represents the fluctuation component of the original sequence, and has a certain periodicity and randomness, which is consistent with the periodic change of the livestock and poultry activity process; the low-frequency component It represents the change trend of the original sequence of livestock and poultry activity process. It can be seen that EMD can decompose the fluctuation components, periodic components and trend components of the livestock and poultry activity process step by step, and each decomposed component contains the same deformation information, which reduces the mutual interference between different feature information to a certain extent, and The variation curve of each component decomposed is smoother than the original livestock and poultry activity deformation sequence curve. It can be seen that the EMD empirical mode decomposition can effectively analyze the deformation data of livestock and poultry activity process under the action of multiple factors. The components obtained by the decomposition include the output of the GM (1, 1) gray prediction model and the establishment of multiple NARX neural network prediction models. forecast. Finally, the prediction results of each component are superimposed to obtain the final fusion prediction result. Case studies show that the proposed fusion prediction results have high prediction accuracy.

2. The present invention adopts the GM(1,1) grey prediction model to predict the low-frequency trend of parameter measurement in the process of animal, livestock and poultry activity with a long time span. The GM(1,1) grey prediction model model can predict the low-frequency trend value of the parameter measurement in the future according to the low-frequency trend value of the parameter measurement. In the original sequence for measuring the low-frequency trend, correspondingly remove a data modeling at the beginning of the sequence, and then carry out the prediction of the low-frequency trend of the prediction parameter measurement. And so on, predict the low frequency trend value of parameter measurement. This method is called an equal-dimensional gray-number complement model, and it enables long-term forecasting. The driver can more accurately grasp the change trend of the low-frequency trend of parameter measurement, and prepare for effectively avoiding the fluctuation of the low-frequency trend of parameter measurement.

3. The present invention adopts the ARIMA prediction model based on the original data of the determined value and the fluctuation value of the parameter measurement to obey the time series distribution, and uses the principle that the change of the determined value and the fluctuation value of the parameter measurement has a certain inertia trend, and integrates trend factors and periodic factors. The original time series variables of the deterministic and fluctuating values of the parameters measured by factors such as random errors and other factors, the non-stationary sequence is transformed into a stationary random sequence with zero mean by methods such as differential data transformation, and the ideal one is selected through repeated identification and model diagnosis. The model performs numerical fitting and prediction of deterministic and fluctuating values of parameter measurements. This method combines the advantages of autoregressive and moving average methods, and has the characteristics of not being bound by data types and having strong applicability.

4. The BAM neural network of the binary connection number of the present invention is a two-layer feedback neural network, which can realize the function of heteroassociative memory; when an input signal is added to one layer, the other layer gets the output. Since the initial mode can act on any layer of the network and information can be propagated in both directions, there is no explicit input or output layer. The learning speed of the BAM neural network model is fast, while the convergence speed of BP learning is slow. The final convergence may be the local minimum point instead of the global minimum point, and the BAM neural network model must reach the energy minimum point; the BAM neural network model has feedback When there is an error in the input, the BAM neural network model can not only output the accurate cause of the failure, but also correct the error of the original input. Therefore, the BAM neural network model is suitable for systems that require correction of erroneous input symptoms. The BAM neural network model utilizes the bidirectional associative storage characteristics of the BAM neural network to improve the processing ability of uncertain information of the predicted value of the material parameter sensor in the inference process.

5. The scientificity and reliability of the classification of livestock and poultry activity state grades of the present invention, the ART2 neural network classifier of the binary connection number of this patent, according to the engineering practice experience of livestock and poultry activity state, through the ART2 neural network of binary connection number The magnitude of the output of 5 different predicted values quantifies the dynamic degree of the activity state of livestock and poultry, and the livestock and poultry are in 5 different states of crawling, walking, standing, lying on the back and lying on the side, so as to realize the dynamic performance of the classification of the activity state of livestock and poultry. and scientific classification.

6. The present invention is aimed at the uncertainty and randomness of the problems such as sensor accuracy error, interference and abnormal measurement value in the process of parameter measurement. The patent of the present invention converts the parameter value measured by the parameter sensor into the binary connection coefficient through the parameter detection module. It can effectively deal with the ambiguity, dynamics and uncertainty of the parameters measured by the parameter sensor, and improve the objectivity and reliability of the parameters detected by the parameter sensor value.

7. The present invention aims at the uncertainty and randomness of the problems such as sensor accuracy error, interference and abnormal measurement value in the process of parameter measurement. The patent of the present invention converts the parameter value measured by the parameter sensor into the binary connection coefficient through the parameter detection module. It can effectively deal with the ambiguity, dynamics and uncertainty of the parameters measured by the parameter sensor, and improve the objectivity and reliability of the parameters detected by the parameter sensor value.

Description of drawings

Fig. 1 is the livestock and poultry sign parameter collection and intelligent prediction platform based on cloud platform of the present invention;

Fig. 2 is the livestock and poultry activity big data prediction subsystem of the present invention;

Fig. 3 is the structure diagram of the parameter detection module of the present invention

4 is a functional diagram of a detection node of the present invention;

5 is a functional diagram of a gateway node of the present invention;

FIG. 6 is a software function diagram of the on-site monitoring terminal of the present invention.

Detailed ways

In conjunction with accompanying drawing 1-6, the technical scheme of the present invention is further described:

1. Design of the overall function of the system

The detection system of the invention realizes the detection of the body temperature and activity information parameters of livestock and poultry and the prediction of the posture of the livestock and poultry. . The cloud platform-based livestock and poultry physical parameter collection and intelligent prediction platform consists of livestock and poultry physical parameters detection node, gateway node, on-site monitoring terminal, cloud platform and mobile app, and realizes between detection nodes and between detection nodes and gateway nodes through ZiGBee technology The detection node sends the detected body temperature and activity parameters of the livestock and poultry to the on-site monitoring terminal and the cloud platform through the gateway node, and the gateway node, the cloud platform, the on-site monitoring terminal and the mobile app realize the exchange of livestock and poultry body temperature and activity information parameters. Two-way transmission. The cloud platform-based livestock and poultry sign parameter collection and intelligent prediction platform is shown in Figure 1.

Second, the design of the detection node

A large number of detection nodes based on the CC2530 self-organizing communication network are used as the temperature and activity information parameter sensing terminals of livestock and poultry, and the detection nodes realize information interaction with the gateway nodes through the self-organizing communication network. The detection node includes sensors for collecting the temperature and activity information parameters of livestock and poultry and the corresponding signal conditioning circuit, STM32 single chip microcomputer and CC2530 module; the software of the detection node mainly realizes self-organizing network communication and the collection and operation of temperature and activity information parameters of livestock and poultry. preprocessing. The software adopts C language programming, which has a high degree of compatibility, which greatly improves the work efficiency of software design and development, and enhances the reliability, readability and portability of the program code. The structure of the detection node is shown in Figure 4.

3. Gateway Node Design

The gateway node includes CC2530 module, NB-IoT module, STM32 microcontroller and RS232 interface. The gateway node communicates with the detection node through the CC2530 module. The NB-IoT module realizes the two-way data interaction between the gateway and the cloud platform, and the RS232 interface connects the scene. The monitoring terminal realizes the information exchange between the gateway and the on-site monitoring terminal. The structure of the gateway node is shown in Figure 5.

4. Software design of on-site monitoring terminal

The on-site monitoring terminal is an industrial control computer. The on-site monitoring terminal mainly realizes the collection of livestock and poultry body temperature and activity information parameters and the prediction of livestock and poultry posture, and realizes the information interaction with the detection node through the gateway node. The main function of the on-site monitoring terminal is communication. Parameter setting, data analysis and data management, and intelligent prediction of livestock and poultry posture through the livestock and poultry activity big data prediction subsystem, the management software selects Microsoft Visual++6.0 as the development tool, and calls the system's Mscomm communication control to design the communication program , the software functions of the on-site monitoring terminal are shown in Figure 6. The structure of the big data prediction subsystem of livestock and poultry activities is shown in Figure 2. The livestock and poultry activity big data prediction subsystem consists of 3 parameter detection modules, BAM neural network model, 2 beat delay line TDL, 2 ARIMA prediction models and ART2 neural network model parts. The function of the livestock and poultry activity big data prediction subsystem is Figure 2 shows the design process as follows:

1. Design of parameter detection module

The three-axis acceleration sensor ADXL362 senses the acceleration of the detected livestock and poultry in the X, Y and Z axis directions as the input of the corresponding parameter detection module, and the three contact numbers output by the three parameter detection modules are used as the input of the BAM neural network model. The output of the neural network model is used as the input corresponding to the ART2 neural network model; the parameter detection module consists of the Adaline neural network model with time delay unit, the EMD empirical mode decomposition model, the GM (1, 1) gray prediction model, and multiple NARX neural networks. The prediction model, 2 beat-by-beat delay lines TDL, 2 ARIMA prediction models and the wavelet neural network model of the binary connection coefficient are composed; the function diagram of the parameter detection module is shown in Figure 3;

(1), Adaline neural network model design with delay unit

The output of the parameter sensor is used as the input of the Adaline neural network model with time delay unit, and the output of the Adaline neural network model with time delay unit is used as the input of the EMD empirical mode decomposition model; the Adaline neural network model with time delay unit is composed of 2 The delay line TDL is composed of an Adaline neural network. The output of the parameter sensor is used as the input of the corresponding beat delay line TDL. The output of the beat delay line TDL is used as the input of the Adaline neural network, and the output of the Adaline neural network is used as the corresponding beat delay line. The input of the TDL, the output of the beat delay line TDL is the output of the Adaline neural network model with a delay unit; the Adaptive Linear Element of the Adaline neural network model is one of the early neural network models. The input signal can be written in the form of a vector: X(K)=[x ₀ (K),x ₁ (K),...x _n (K)] ^T , each set of input signals corresponds to a set of weight vectors correspondingly expressed as :W(K)=[k ₀ (K),k ₁ (K),…k(K)], when x ₀ (K) is equal to negative 1, the bias value of the Adaline neural network model determines the excitation of neurons or Inhibition state, the network output can be defined according to the input vector and weight vector of the Adaline neural network model as:

In the Adaline neural network model, there is a special input, the ideal response output d(K), which is sent to the Adaline neural network model, and then compared through the network output y(K), and the difference is sent to the learning algorithm mechanism. , to adjust the weight vector until the best weight vector is obtained, y(K) and d(K) tend to be consistent, the adjustment process of the weight vector is the learning process of the network, and the learning algorithm is the core part of the learning process. The weight optimization search algorithm adopts the least square method of LMS algorithm.

(2), EMD empirical mode decomposition model design

The parameter sensor output is used as the input of the Adaline neural network model with time-delay unit, the output of the Adaline neural network model with time-delay unit is used as the input of the EMD empirical mode decomposition model, and the parameter measurement low-frequency trend value output by the EMD empirical mode decomposition model is used as The input of the GM (1, 1) gray prediction model, and the high-frequency fluctuation values of multiple parameters output by the EMD empirical mode decomposition model are respectively used as the input of the corresponding multiple NARX neural network prediction models. The EMD empirical mode decomposition is a kind of The adaptive signal screening method has the characteristics of simple calculation, intuitive, experience-based and self-adaptive. It can screen out the trends of different features in the parameter measurement signal step by step, and obtain multiple high frequency fluctuation parts (IMF) and low frequency trend parts. The IMF components decomposed by the EMD empirical mode contain the components of different frequency segments of the parameter measurement signal from high to low. The frequency resolution contained in each frequency segment varies with the signal itself, and has the characteristics of adaptive multi-resolution analysis. The purpose of using EMD empirical mode decomposition is to extract parameter measurement information more accurately. The IMF component must satisfy two conditions at the same time: (1) In the parameter measurement signal to be decomposed, the number of extreme points is equal to the number of zero-crossing points, or differs by at most one; (2) At any time, the local maximum and the local minimum The value defines the envelope mean of zero. The steps of the "sieving" process of the EMD empirical mode decomposition method for the output value signal of the Adaline neural network model with time-delay units are as follows:

(a) The Adaline neural network model with time-delay unit outputs all the local extremum points of the value signal, and then uses a cubic spline to connect the left and right local maxima points to form an upper envelope.

(b) The lower envelope is formed by connecting the local minimum points of the output value of the Adaline neural network model with time-delay units with a cubic spline, and the upper and lower envelopes should envelop all the data points.

(c) The average value of the upper and lower envelopes is denoted as m ₁ (t), and we can obtain:

x(t)-m ₁ (t)=h ₁ (t) (2)

x(t) is the original signal of the output value of the Adaline neural network model with delay unit. If h ₁ (t) is an IMF, then h ₁ (t) is the first IMF component of x(t). Denote c ₁ (t)=h _1k (t), then c ₁ (t) is the first component of the signal x(t) that satisfies the IMF condition.

(d) Separating c ₁ (t) from x(t) yields:

r ₁ (t)=x(t)-c ₁ (t) (3)

Repeat steps (a) to (c) with r ₁ (t) as the original data to obtain the second component c ₂ of x(t) that satisfies the IMF condition. Repeat the cycle n times to obtain n components of the signal x(t) that satisfy the IMF condition. In this way, the output of the Adaline neural network model with time-delay unit is decomposed into low-frequency trend parts and multiple high-frequency fluctuation parts through the EMD empirical mode decomposition model. The EMD empirical decomposition model is shown in Figure 2.

(3), GM (1, 1) grey prediction model design

The measured low-frequency trend value of the parameters output by the EMD empirical mode decomposition model is used as the input of the GM(1,1) gray prediction model, and the measured high-frequency fluctuation values of multiple parameters output by the EMD empirical mode decomposition model are respectively used as the corresponding multiple NARX neural The input of the network prediction model, the output of the GM(1,1) gray prediction model and the outputs of multiple NARX neural network prediction models are respectively used as the corresponding input of the wavelet neural network model of the binary connection coefficient; the GM(1,1) gray prediction method is relatively The traditional statistical prediction method has many advantages. It does not need to determine whether the predictor variable obeys the normal distribution, does not require large sample statistics, and does not need to change the prediction model at any time according to the change of the input variable of the low-frequency trend value measured by the parameter. The cumulative generation technology establishes a unified differential equation model, and the cumulative parameter measurement low-frequency trend original value is restored to obtain the prediction result, and the differential equation model has higher prediction accuracy. The essence of establishing the GM(1,1) grey prediction model is to accumulate and generate the original data of the low-frequency trend value once, so that the generated sequence shows a certain law. By establishing the differential equation model, the fitting curve is obtained to measure the low-frequency trend of the parameters. value to predict.

(4), NARX neural network prediction model design

The measured low-frequency trend value of the parameters output by the EMD empirical mode decomposition model is used as the input of the GM(1,1) gray prediction model, and the measured high-frequency fluctuation values of multiple parameters output by the EMD empirical mode decomposition model are respectively used as the corresponding multiple NARX neural The input of the network prediction model, the output of the GM (1, 1) gray prediction model and the outputs of multiple NARX neural network prediction models are respectively used as the corresponding input of the wavelet neural network model of the binary connection coefficient; the NARX neural network prediction model is a kind of with output The dynamic recurrent neural network with feedback connection can be equivalent to a BP neural network with input delay plus a delay feedback connection from output to input in terms of topological connection relationship. Its structure consists of input layer, delay layer, hidden layer and output. Layer composition, in which the input layer node is used for signal input, the delay layer node is used for the time delay of the input signal and the output feedback signal, the hidden layer node uses the activation function to perform nonlinear operations on the delayed signal, and the output layer node uses The final network output is obtained by linearly weighting the output of the hidden layer. The output h _i of the ith hidden layer node of the NARX neural network prediction model is:

The output o _j of the jth output layer node of the NARX neural network is:

The input layer, the delay layer, the hidden layer and the output layer of the NARX neural network of the patent of the present invention are respectively 2-19-10-1 nodes.

(5), ARIMA prediction model design

The parameter measurement determination value a and the fluctuation value b output by the wavelet neural network of the binary connection number are used as the input of the corresponding beat delay line TDL, respectively, and the two beat delay line TDL outputs are used as the input of the corresponding ARIMA prediction model, 2 The outputs of each ARIMA prediction model are respectively used as the corresponding input of the wavelet neural network model of the binary connection coefficient. The model (Moving Average, MA) is organically combined to make it a comprehensive forecasting method. As one of the effective modern data processing methods, it is known as the most complex and advanced model in time series forecasting methods. In practical applications, because the input original data series often show a certain trend or cycle characteristics, it does not satisfy the ARMA model. The stationarity of time series is required, and taking the difference is a convenient and effective method to eliminate the trend of the data. The model established based on the differenced data sequence is called the ARIMA model, denoted as {Xt}-ARIMA(p, d, q), where p and q are called the order of the model, and d represents the number of differences. Obviously, when d is 0, the ARIMA model is an ARMA model, which is defined as:

x _t = b ₁ x _t-1 +…+b _p x _tp +ε _t +a ₁ ε _t-1 +…+a _q ε _tq (6)

{x _t } is the data sequence of the parameter measurement determination value a and the fluctuation value b output by the wavelet neural network of the binary connection coefficient to be predicted, {ε _t }～WN(0,σ ² ). ARIMA model establishment mainly includes model identification, parameter estimation and model diagnosis. Model identification mainly includes time series preprocessing and initial order determination of model parameters; after the model order determination is completed, it is necessary to estimate the unknown parameters in the model through time series observations combined with p, d, and q values; model diagnosis is mainly is the significance test for the entire model and the significance test for the parameters in the model. Usually the establishment of the model is a process of continuous optimization. The commonly used model optimization is the AIC and BIC criteria, that is, the smaller the value of the minimum information criterion, the more suitable the model is. The BIC criterion is an improvement for the deficiency of the AIC criterion for large sample sequences. . The ARIMA(p,d,q) model can be used to fit the time series. The ARIMA(p,d,q) modeling steps are as follows:

A. Obtain the parameter measurement determination value a and the fluctuation value b sequence output by the wavelet neural network of the binary connection coefficient.

B. Judging the stationarity of the sequence, if the sequence is not stationary, it is necessary to perform data preprocessing and difference operation on the data to make the sequence stationary, and determine the value of the difference order d.

C. When the sequence after difference is a stationary non-white noise sequence, we can choose an ARMA(p,q) model with appropriate order to model the sequence.

D. Estimate the unknown parameters in the model according to the identified model and its order.

E. Test the residual sequence, and use the method of statistical test to test whether the preliminary model is valid.

F. Use the obtained fitting model to predict the future development trend of the stationary time series.

(6) The wavelet neural network model design of binary connection number

The wavelet neural network model of the binary connection number outputs the determined value a and the fluctuation value b of the parameter measurement value. The binary connection number of the parameter measurement value is a+bi, and the determined value a and the fluctuation value b of the parameter measurement value are taken as the corresponding The input of the beat delay line TDL and the two corresponding inputs of the wavelet neural network model of the binary connection number, the two beat delay line TDL outputs are respectively used as the input of the corresponding ARIMA prediction model, and the outputs of the two ARIMA prediction models are respectively used as two The corresponding input of the wavelet neural network model of the element connection number, the output of the wavelet neural network model of the binary connection number is the binary connection value of the parameter output by the parameter detection module; the structure of the parameter detection module is shown in Figure 2. The wavelet neural network model WNN (Wavelet Neural Networks) is a feedforward network proposed by combining artificial neural network on the basis of wavelet theory. It uses the wavelet function as the excitation function of the neuron, and the wavelet scaling, translation factor and connection weight are adaptively adjusted in the process of optimizing the error energy function. Assuming that the input signal of the wavelet neural network model can be represented as an input one-dimensional vector x _i (i=1,2,…,n), and the output signal is represented as _yk (k=1,2,…,m), the wavelet neural network The calculation formula of the output value of the output layer of the network model is:

为小波基函数，b_j为小波基函数的平移因子，a_j小波基函数的伸缩因子，ω_jk为隐含层j节点和输出层k节点间的连接权值。本专利中的小波神经网络模型的权值和阈值的修正算法采用梯度修正法来更新网络权值和小波基函数参数，从而使小波神经网络输出不断逼近期望输出。小波神经网络模型的输出为代表一段时间参数测量传感器值大小的动态二元联系数，动态二元联系数为a+bi,a+bi构成在一段时间参数测量传感器输出的被测量参数的动态二元联系数值。In the formula, ω _ij is the connection weight between the input layer i node and the hidden layer j node,

is the wavelet basis function, b _j is the translation factor of the wavelet basis function, a _j is the scaling factor of the wavelet basis function, ω _jk is the connection weight between the hidden layer j node and the output layer k node. The correction algorithm for the weights and thresholds of the wavelet neural network model in this patent uses the gradient correction method to update the network weights and wavelet basis function parameters, so that the output of the wavelet neural network is constantly approaching the desired output. The output of the wavelet neural network model is a dynamic binary connection number representing the magnitude of the sensor value of the parameter measurement sensor for a period of time. Meta contact value.

2. BAM neural network model design

The three connection numbers output by the three parameter detection modules are used as the input of the BAM neural network model, and the binary connection number output by the BAM neural network model is used as the input corresponding to the ART2 neural network model; in the topology of the BAM neural network model, the initial value of the network input The mode is x(t), which is weighted by the weight matrix W ₁ to reach the output terminal y, and then returned to the input terminal x after the nonlinear transformation of the transfer characteristic f _y of the output node and W ₂ matrix weighting, and then output through the x terminal The nonlinear transformation of the node transfer characteristic f _x becomes the output of the input terminal x, and this operation process is repeated. The state transition equation of the BAM neural network model is shown in equation (8).

The output of the BAM neural network model is the dynamic binary connection number representing the activity state of livestock and poultry for a period of time.

3. Design of ARIMA prediction model

The determined value and the fluctuation value output by the ART2 neural network model are used as the input of the two beat delay line TDLs and the corresponding input of the ART2 neural network model, and the two beat delay line TDL outputs are respectively used as the corresponding two ARIMA prediction model inputs, 2 The output of each ARIMA prediction model is used as the corresponding input of the ART2 neural network model; the design method of the ARIMA prediction model refers to the design process and method of the ARIMA prediction model of the parameter detection module.

4. Design of ART2 neural network model

The binary connection number output by the BAM neural network model is used as the input corresponding to the ART2 neural network model, the definite value and the fluctuation value output by the ART2 neural network model are used as the input of the two beat-by-beat delay lines TDL and the corresponding input of the ART2 neural network model, 2 The TDL outputs of each beat delay line are respectively used as the input of the corresponding two ARIMA prediction models, and the outputs of the two ARIMA prediction models are used as the corresponding inputs of the ART2 neural network model. The 5 different binary connection numbers correspond to 5 different states of the livestock and poultry in recumbency, walking, standing, supine and side recumbency; the ART2 neural network model structure includes the F1 attention subsystem and the F2 orientation subsystem, and the F1 layer is divided into superior,

The middle and lower layers, of which the lower layer and the middle layer, the middle layer and the upper layer respectively form two closed positive feedback loops to realize the functions of feature enhancement and noise suppression; it is composed of long-term and short-term memory. There are M neurons in F1 and F2, which constitute The M-dimensional state vector represents the short-term memory STM of the network; the internal and external connection weight vectors of F1 and F2 constitute the adaptive long-term memory LTM of the network; the network contains two kinds of functional neurons, which are represented by hollow and solid, among which hollow neurons represents the input excitation superposition, and the solid neurons represent the norm of the input vector. The definite value c and the fluctuation value d output by the ART2 neural network model, they constitute a binary connection number c+di, and the five different binary connection numbers correspond to the animals and poultry in recumbent crawling, walking, standing, supine and side recumbency. 5 different states. The corresponding relationship between the binary connection number output by the ART2 neural network model and the state of livestock and poultry is shown in Table 1.

Table 1 Correspondence table of five different binary connection numbers and livestock and poultry status

5. Cloud-based platform design of livestock and poultry sign parameter collection and intelligent prediction platform

It consists of detection node, gateway node, on-site monitoring terminal, cloud platform and mobile app for the physical parameters of livestock and poultry. The communication between detection nodes and between detection nodes and gateway nodes is realized through ZiGBee technology; the detection node will detect the body temperature and activity of livestock and poultry. The parameters are sent to the on-site monitoring terminal and the cloud platform through the gateway node, and the two-way transmission of livestock and poultry body temperature and activity information parameters between the gateway node, cloud platform, on-site monitoring terminal and mobile app is realized; according to the distribution of livestock and poultry parameters, the wearing method is adopted. The detection node is worn on the body surface of the livestock and poultry, and the gateway node and the on-site monitoring terminal are placed in the livestock and poultry farm. The detection node realizes the detection of the temperature and activity parameter information of the livestock and poultry. Intelligent prediction of bird posture.

The technical means disclosed in the solution of the present invention are not limited to the technical means disclosed in the above embodiments, but also include technical solutions composed of any combination of the above technical features. It should be pointed out that for those skilled in the art, without departing from the principle of the present invention, several improvements and modifications can also be made. These improvements and modifications are also regarded as the protection scope of the present invention.

Claims (7)

1. A big data detection system for livestock and poultry activity information, characterized in that: the detection system includes a cloud platform-based livestock and poultry sign parameter collection and intelligent prediction platform and a livestock and poultry activity big data prediction subsystem, and a cloud platform-based livestock and poultry The physical sign parameter collection and intelligent prediction platform realizes the detection and processing of temperature and activity information parameters of livestock and poultry signs; the livestock and poultry activity big data prediction subsystem realizes the prediction of the activity state of livestock and poultry, and provides data and early warning for the prevention of livestock and poultry diseases; The livestock and poultry activity big data prediction subsystem includes a parameter detection module, a BAM neural network model, a beat delay line TDL, an ARIMA prediction model and an ART2 neural network model, and a three-axis acceleration sensor senses the X, Y and Z of the livestock and poultry to be detected. The acceleration in the axial direction is used as the input of the corresponding parameter detection module, the three connection numbers output by the three parameter detection modules are used as the input of the BAM neural network model, and the binary connection number output by the BAM neural network model is used as the corresponding ART2 neural network model. Input, the determined value and fluctuation value of the output of the ART2 neural network model are used as the input of the corresponding beat delay line TDL and the corresponding input of the ART2 neural network model, and the output of the beat delay line TDL is used as the input of the corresponding ARIMA prediction model, and the output of the ARIMA prediction model As the corresponding input of the ART2 neural network model, the definite value c and the fluctuation value d output by the ART2 neural network model constitute a binary connection number c+di, and the binary connection number output by the ART2 neural network model corresponds to the different states of livestock and poultry. 2. the big data detection system of livestock and poultry activity information according to claim 1, is characterized in that: described parameter detection module is made up of Adaline neural network model with time-delay unit, EMD empirical mode decomposition model, GM (1, 1) The gray prediction model, the NARX neural network prediction model, the beat delay line TDL, the ARIMA prediction model and the wavelet neural network model of the binary connection coefficient are composed. 3. the big data detection system of livestock and poultry activity information according to claim 2 is characterized in that: the parameter detection sensor output is as the input of the Adaline neural network model with time-delay unit, and the Adaline neural network model output with time-delay unit As the input of the EMD empirical mode decomposition model, the low-frequency trend value of the measured parameters output by the EMD empirical mode decomposition model is used as the input of the GM (1, 1) gray prediction model, and the EMD empirical mode decomposition model outputs many measured parameters. Different high-frequency fluctuation values are used as the input of the corresponding multiple NARX neural network prediction models, and the output of the GM (1, 1) gray prediction model and the output of multiple NARX neural network prediction models are respectively used as the wavelet neural network model of the binary connection coefficient. Corresponding to the input, the wavelet neural network model of the binary connection number outputs the determined value a and the fluctuation value b of the measured parameter to form the binary connection number a+bi of the measured parameter size, the determined value a and the fluctuation value b of the measured parameter are respectively As the input of the corresponding beat delay line TDL and the corresponding input of the wavelet neural network model of the binary connection number, the output of the beat delay line TDL is respectively used as the input of the corresponding ARIMA prediction model, and the output of the ARIMA prediction model is respectively used as the binary connection number. The corresponding input of the wavelet neural network model, the binary connection number output by the wavelet neural network model of the binary connection number is used as the predicted value of the measured parameter. 4 . The big data detection system for livestock and poultry activity information according to claim 1 , wherein the different states include crawling, walking, standing, lying on the back and lying on the side. 5 . 5. The big data detection system for livestock and poultry activity information according to claim 1, characterized in that: the cloud platform-based livestock and poultry physical parameter collection and intelligent prediction platform comprises a detection node, a gateway node, an on-site monitoring terminal, a cloud platform platform and mobile APP. 6. The big data detection system for livestock and poultry activity information according to claim 5, wherein the detection node collects the body temperature and activity information parameters of livestock and poultry and uploads it to the cloud platform through the gateway node, and stores data on the cloud platform side The mobile APP can monitor the physical signs and motion parameters of livestock and poultry in real time through the temperature and activity information of livestock and poultry provided by the cloud platform. The detection node is responsible for collecting the body temperature and activity information parameters of livestock and poultry. , two-way communication between cloud platform and mobile APP to realize the collection and prediction of body temperature and activity information parameters of livestock and poultry. 7 . The big data detection system for livestock and poultry activity information according to claim 1 , wherein the model of the three-axis acceleration sensor is ADXL362. 8 .

Images