CN116029604B - A method of regulating the caged meat duck breeding environment based on health and comfort - Google Patents

Abstract

The invention discloses a method for regulating and controlling the breeding environment of cage-raised meat ducks based on health comfort, which belongs to the technical field of environment regulation and control and comprises the following steps: s1: acquiring monitoring information of the internal and external environments of a duck shed and health characterization information of meat ducks; s2: determining key indexes of health comfort level of the meat ducks according to the internal and external environment monitoring information of the duck shed and the health characterization information of the meat ducks; s3: determining the health comfort index of the meat ducks according to the health comfort key index of the meat ducks; s4: taking the internal and external environment monitoring information of the duck shed and the health comfort index of the meat duck as inputs of a neural network to obtain the internal environment information of the duck shed and the health comfort index predicted value of the meat duck at the next moment; s5: and carrying out environment regulation according to the environmental information in the duck shed and the meat duck health comfort index predicted value at the next moment. Compared with the prior art, the method takes the external environment information and meat duck information of the breeding device into consideration, so that the environment regulation and control reference information is more comprehensive and scientific, and the accuracy and reliability of the environment regulation and control are improved.

Description

A method of regulating the caged meat duck breeding environment based on health and comfort

Technical field

The invention belongs to the technical field of environmental regulation, and specifically relates to a caged meat duck breeding environment regulation method based on health and comfort.

Background technique

As the technology and equipment of the poultry breeding industry continue to improve, the caged meat duck breeding method has developed rapidly. Due to the high breeding density of caged meat ducks and the limited range of activities of meat ducks, the environmental conditions of the breeding houses have become a key constraint affecting the health and production performance of meat ducks. Therefore, it is of great significance to control the caged meat duck breeding environment to be within a suitable range.

At present, the environmental control system of poultry breeding houses is a control method based on the monitoring results of environmental parameters. The monitoring data of environmental sensors in the breeding house are used to determine whether it meets the requirements of the indoor environment for control. Zhang Longai and others designed an air conditioning control method, control device and system for breeding, which can automatically control the air environment of the breeding place, but can only regulate the air and air volume. Zhou Feng et al. proposed an intelligent multi-layer three-dimensional meat duck breeding system based on big data, which includes environmental monitoring and environmental regulation modules. Environmental regulation is carried out based on detected environmental changes. The adjustment lag of environmental regulation equipment is not considered, resulting in environmental regulation. Insufficient real-time and accuracy. Li Baoming and others proposed a livestock and poultry house breeding environment temperature prediction control system and its regulation method. Based on the temperature prediction model, the temperature of the breeding environment is fuzzy controlled, which improves the real-time and adaptability of temperature control in the breeding house. However, it does not Optimizing the temperature control model for animal comfort is lacking in scientificity and adaptability. Therefore, a method that integrates multiple environmental parameters and meat duck monitoring information is needed to improve the real-time, accuracy, scientificity and adaptability of environmental control in breeding houses.

Contents of the invention

The purpose of the present invention is to solve the problems of insufficient scientificity and adaptability and low real-timeness and accuracy of existing environmental regulation, and proposes a method for regulating the environment of caged meat duck breeding based on health and comfort.

The technical solution of the present invention is: a caged meat duck breeding environment control method based on health and comfort includes the following steps:

S1: Obtain environmental monitoring information inside and outside the duck house and meat duck health characterization information;

S2: Determine the key indicators of health and comfort of meat ducks based on environmental monitoring information inside and outside the duck house and meat duck health characterization information;

S3: Determine the meat duck health and comfort index based on the key indicators of meat duck health and comfort;

S4: Use the environmental monitoring information inside and outside the duck house and the health and comfort index of meat ducks as the input of the neural network to obtain the predicted value of the environmental information inside the duck house and the health and comfort index of meat ducks at the next moment;

S5: Carry out environmental regulation based on the environmental information in the duck house at the next moment and the predicted value of the meat duck health and comfort index.

Further, in step S1, the meat duck health representation information includes abnormal behavior awareness information, abnormal feed intake awareness information, abnormal movement awareness information, abnormal sound awareness information and abnormal body temperature awareness information;

The specific method of obtaining abnormal behavior awareness information and abnormal feed intake awareness information is: collecting duck house videos, and using the adaptive threshold method to obtain meat duck images from the duck house videos, and using the meat duck images as the input of the convolutional neural network. Obtain information on abnormal behavior and abnormal feed intake;

The specific method of obtaining sound abnormality awareness information is: collecting meat duck sound data, sequentially performing pre-emphasis, framing, windowing, Fourier transform and Mel frequency cepstrum coefficient solution on the meat duck sound data to obtain the sound feature vector, As sound abnormality awareness information;

The specific method to obtain motion abnormality awareness information is: collect meat duck motion data, use signal time domain analysis method, signal frequency domain analysis method and signal time-frequency domain analysis method to extract multi-dimensional features of meat duck motion data, obtain the motion feature vector, and The sound feature vector and the motion feature vector are fused to obtain a fused feature vector, which is used as motion abnormality awareness information;

The specific method of obtaining the abnormal body temperature awareness information is to collect the body temperature data of meat ducks, and use the body temperature data outside the set temperature threshold range as the abnormal body temperature awareness information.

Further, step S2 includes the following sub-steps:

S21: Sequentially pair the environmental monitoring information inside and outside the duck house with the meat duck health characterization information and add time information. Use the collection of environmental monitoring information inside and outside the duck house as the environmental monitoring transaction database, and use the collection of meat duck health characterization information as the meat duck The health characterization transaction database uses the collection of environmental monitoring transaction database, meat duck health characterization transaction database and time information as the general transaction database;

S22: Calculate transaction support and transaction confidence based on the total transaction database;

S23: Calculate the transaction improvement degree based on the transaction support and transaction confidence;

S24: Use the data whose transaction improvement degree is greater than the set improvement threshold as a key indicator of the health and comfort of meat ducks.

Further, in step S22, the calculation formula of transaction support Support(X _i ,Y _i ) is:

In the formula, P(X _i ,Y _i ) represents the proportion of transactions containing the environmental monitoring information inside and outside the duck house in the X _i state and the meat duck health characterization information in the Y _i state to the total transaction library, number(X _i ,Y _i ) Represents the number of transactions containing environmental monitoring information inside and outside the duck house in the X _i state and meat duck health characterization information in the Y _i state, and number(ALLSamples) represents the number of transactions in the total transaction library;

In step S22, the calculation formula of the transaction confidence level Confidence (X _i →Y _i ) is:

In _{the formula , P (} The probability that the internal and external environment monitoring information is in the X _i state;

In step S23, the calculation formula of the transaction lifting degree Lift(X _i ,Y _i ) is:

In the formula, P(Y _i ) represents the probability that the meat duck health representation information is in the Y _i state.

Further, step S3 includes the following sub-steps:

S31: Use the environmental monitoring information inside and outside the duck house and the meat duck health characterization information as the element layer, use the key indicators of meat duck health and comfort as the indicator layer, and calculate the subjective weight based on the element layer and the indicator layer;

S32: Calculate the objective weight based on the element layer and indicator layer;

S33: Calculate comprehensive subjective and objective weights based on subjective weights and objective weights;

S34: Calculate the meat duck health and comfort index based on the comprehensive subjective weight.

Further, in step S31, the calculation formula of subjective weight μ ¹ is:

In the formula, U represents the number of elements in the feature layer, C represents the number of elements in the indicator layer, and ω represents the weight of each element;

In step S32, the key indicators of meat duck health and comfort are normalized, the comprehensive data value of the i-th indicator after normalization is calculated, the coefficient of variation is calculated based on the comprehensive data value of the i-th indicator, and the coefficient of variation is calculated. Objective weight; among them, the comprehensive data value of the i-th indicator The calculation formulas of the coefficient of variation b _i and objective weight μ ² are respectively:

In the formula, t represents the number of key indicators, Z _il represents the l-th data of the i-th key indicator, x _ij represents the parameter value of the i-th indicator belonging to the j-th type of key indicators among the t key indicators, Represents the comprehensive data value of the jth indicator;

In step S33, the calculation formula of the comprehensive subjective and objective weight β _i is:

In the formula, α _i represents the empirical factor of the i-th indicator, is the subjective weight of the i-th indicator,/> is the objective weight of the i-th indicator;

In step S34, the calculation formula of the meat duck health and comfort index is:

In the formula, Z _i represents the i-th key indicator.

Further, in step S5, the specific method for environmental control is: calculate the deviation and deviation change rate between the environmental information in the duck house and the preset environmental information at the next moment. When the deviation or the deviation change rate is greater than the set threshold, the meat is The predicted value of duck health and comfort index is used as a constraint for environmental regulation.

The beneficial effects of the present invention are:

(1) The present invention collaboratively obtains environmental monitoring information and meat duck monitoring information, performs health awareness processing on the meat duck monitoring information to obtain meat duck health characterization information, and combines the correlation analysis method to efficiently integrate multi-source information, which is consistent with the existing technology. Compared with the above, the external environmental information of the breeding facility and the meat duck information are taken into consideration, making the environmental control reference information more comprehensive and scientific, and improving the accuracy and reliability of environmental control;

(2) The meat duck health and comfort evaluation process used in the present invention is an upgrade of the existing comprehensive environmental assessment system for breeding houses. The health characterization information of meat ducks is included in the evaluation indicators. The integration of subjective and objective weights makes the weight of each indicator more scientific and rigorous. , improving the accuracy and reliability of dynamic prediction of environmental information and health and comfort;

(3) The meat duck breeding environment control method of the present invention is formulated based on the prediction of environmental parameters and with the meat duck health and comfort prediction index as a constraint. Compared with the existing control method based on environmental parameter detection results, it is more scientific. Accurate and more real-time.

Description of the drawings

Figure 1 is a flow chart of the environment control method for caged meat duck breeding;

Figure 2 is a structural diagram of the system corresponding to the environment control method for caged meat duck breeding.

Detailed ways

The embodiments of the present invention will be further described below with reference to the accompanying drawings.

As shown in Figure 1, the present invention provides a caged meat duck breeding environment control method based on health and comfort, which includes the following steps:

S1: Obtain environmental monitoring information inside and outside the duck house and meat duck health characterization information;

S2: Determine the key indicators of health and comfort of meat ducks based on environmental monitoring information inside and outside the duck house and meat duck health characterization information;

S3: Determine the meat duck health and comfort index based on the key indicators of meat duck health and comfort;

S4: Use the environmental monitoring information inside and outside the duck house and the health and comfort index of meat ducks as the input of the neural network to obtain the predicted value of the environmental information inside the duck house and the health and comfort index of meat ducks at the next moment;

S5: Carry out environmental regulation based on the environmental information in the duck house at the next moment and the predicted value of the meat duck health and comfort index.

In step S1, health awareness processing is performed on the meat duck monitoring information according to the time sequence to obtain the meat duck health representation information (feed intake, movement trajectory, behavior, etc.), and the environmental information and the meat duck health representation information are paired in time series and time information is added. .

In step S2, the environmental monitoring information and meat duck health characterization information obtained in step S1 are data fused based on time information, and then a correlation analysis algorithm is used to mine the relationship between the fused data and find out the strong correlation rules between multi-source information. , extracting key indicators as important indicators for the establishment of meat duck health and comfort evaluation system.

In step S3, the important indicators obtained in step S2 are input into the meat duck health and comfort evaluation system to obtain the current meat duck health and comfort index. The environmental information and the meat duck health and comfort are time-series matched and time information is added.

In step S4, the neural network technology is used to establish a prediction model, and the environmental monitoring information and the meat duck health and comfort index time series are input into the neural network to obtain the environmental information and the dynamic prediction value of the meat duck health and comfort index in the next time period.

In step S5, the predicted value of the environmental information obtained in step S4 is compared with the appropriate environmental information to obtain the deviation and the deviation change rate. The deviation and deviation change rate of the environmental information prediction value are input into the fuzzy control model to obtain the fuzzy control amount of each environmental information in the farm. The predicted value of the health and comfort of the meat ducks is used as a constraint, and the environmental regulation equipment regulation rules are formulated in combination with the qualitative decision-making method to form a cage. Decision-making on environmental regulation of meat duck breeding.

In the embodiment of the present invention, in step S1, the meat duck health representation information includes abnormal behavior awareness information, abnormal feed intake awareness information, abnormal movement awareness information, abnormal sound awareness information and abnormal body temperature awareness information;

The specific method of obtaining abnormal behavior awareness information and abnormal feed intake awareness information is: collecting duck house videos, and using the adaptive threshold method to obtain meat duck images from the duck house videos, and using the meat duck images as the input of the convolutional neural network. Obtain information on abnormal behavior and abnormal feed intake;

The specific method of obtaining sound abnormality awareness information is: collecting meat duck sound data, sequentially performing pre-emphasis, framing, windowing, Fourier transform and Mel frequency cepstrum coefficient solution on the meat duck sound data to obtain the sound feature vector, As sound abnormality awareness information;

The specific method to obtain motion abnormality awareness information is: collect meat duck motion data, use signal time domain analysis method, signal frequency domain analysis method and signal time-frequency domain analysis method to extract multi-dimensional features of meat duck motion data, obtain the motion feature vector, and The sound feature vector and the motion feature vector are fused to obtain a fused feature vector, which is used as motion abnormality awareness information;

Comprehensive use of signal time domain, frequency domain, and time-frequency domain analysis methods to extract multi-dimensional features and construct motion feature vectors that characterize meat duck behavior. A deep fusion network is constructed by stacking autoencoders to alternately optimize and fuse sound features and motion features to obtain sound. -Motion fusion feature vector, applying domain adaptation theory to obtain motion abnormality awareness information;

The specific method of obtaining the abnormal body temperature awareness information is to collect the body temperature data of meat ducks, and use the body temperature data outside the set temperature threshold range as the abnormal body temperature awareness information.

For the accessed video stream data, the adaptive threshold method is applied to segment the meat duck image to quickly and accurately extract the meat duck image area. On the basis of meat duck image segmentation, the Kalman filter algorithm is combined with the meat duck target tracking. First, the proposed adaptive threshold meat duck image segmentation algorithm is used to find the complete moving target area, traverse each contour, and calculate the centroid point coordinates. An improved algorithm combining Kalman filtering and foreground target segmentation algorithm is used for real-time tracking to achieve dynamic meat duck image segmentation. Behavioral tracking of duck targets. Based on the convolutional neural network of deep learning, train and establish a convolutional artificial neural network model for the meat duck cage environment, and obtain information such as abnormal behavior awareness, abnormal feed intake awareness, and abnormal movement awareness of meat ducks.

Amplify the incoming meat duck sound data, analyze the characteristics of the collected meat duck sounds, select an appropriate filtering method, and determine the location of the meat ducks that emit abnormal sounds through a combination of frequency identification and positioning algorithms. Based on MFCC (Mel Frequency Cepstrum Coefficient), acoustic features suitable for meat duck sound data are designed. Through data pre-emphasis, framing, windowing, fast Fourier transform and MFCC solution, the sound feature vector that characterizes meat duck behavior is obtained. Using meat duck movement behavior data, the fluctuation of the curve of meat duck movement behavior in the X, Y, and Z axes is analyzed. The signal time domain, frequency domain and time-frequency domain analysis methods are comprehensively used to extract multi-dimensional features to form a representation of meat duck behavior. motion feature vector. In order to give full play to the correlation and complementarity of multi-source heterogeneous data, a deep fusion network is constructed by stacking autoencoders to alternately optimize the fusion of sound features and motion features, and obtain a sound-motion fusion feature vector. In view of the problem of category imbalance in the data set collected by the sensor, that is, the number of some categories is much more than the number of other categories, applying the domain adaptation theory, first normalize the fusion feature vector, define the boundary loss function, and automatically Adaptively combine known behavioral classifications to determine the center and radius of the spherical decision boundary, and then determine the decision boundary for meat duck behavior recognition, and obtain information such as abnormal behavior awareness, abnormal movement awareness, and abnormal sound awareness.

Comprehensive information such as abnormal behavior awareness, abnormal feed intake awareness, abnormal movement awareness, abnormal voice awareness, and abnormal body temperature awareness are used as meat duck health representation information.

In the embodiment of the present invention, step S2 includes the following sub-steps:

S21: Use the collection of environmental monitoring information inside and outside the duck house as the environmental monitoring transaction database, and use the collection of meat duck health characterization information as the meat duck health characterization transaction database;

S22: Calculate transaction support and transaction confidence based on the environmental monitoring transaction database and the duck health characterization transaction database;

S23: Calculate the transaction improvement degree based on the transaction support and transaction confidence;

S24: Use the data whose transaction improvement degree is greater than the set improvement threshold as a key indicator of the health and comfort of meat ducks.

The multi-source information correlation analysis algorithm based on Apriori designed by the present invention is an application in step S2. This algorithm uses multi-source information for environmental monitoring and meat duck health characterization, and uses an iterative method of layer-by-layer search to extract key indicators to establish a meat duck health and comfort evaluation system. The relevant concepts of the algorithm are as follows:

Transaction: In the present invention _, the data sets Xi and _Yi composed of the environmental monitoring data x _i and the meat duck health characterization data _yi respectively with time information are called transactions, and the data set composed of the transactions Xi _{and Yi} is the environment Monitoring transaction library X and meat duck health characterization transaction library Y. Divide the environmental monitoring information into three ranges: normal, high and low, then one of the transactions X _i can be expressed as:

[Suitable temperature, suitable humidity, high CO2 concentration...]. Divide meat duck health representation information into three levels: normal, healthy and critical, then one of the transactions _Yi can be expressed as: [normal behavior, abnormal movement, normal voice, normal body temperature].

k-itemset: If the transaction _Xi contains k elements, then the _transaction

In the embodiment of the present invention, in step S22, the support degree refers to the probability that several transactions in the transaction library appear at the same time; the calculation formula of the transaction support degree Support (X _i , Yi ₎ is:

In the formula, P(X _i ,Y _i ) represents the proportion of transactions _in the transaction library that include environmental monitoring information _Xi inside and outside the duck house and meat duck _health characterization information _Yi The number of transactions for environmental monitoring information _Xi and meat duck health characterization information _Yi , number(ALLSamples) represents the total number of transactions in the two transaction libraries;

In step S22, the confidence pointer is defined _for the association rule. P(X _{i 〡Y i} ) refers to the _probability that _when transaction _Xi occurs, transaction The calculation formula of Confidence(X _i →Y _i ) is:

In the formula, P(X _i 〡Y _i ) represents the probability that the meat duck health representation information is _{Y i} when the environmental monitoring information is Xi; P(X _i ) represents the probability that the environmental monitoring information is _Xi ;

In step S23, the lift degree refers to the ratio of the confidence level of "X _i → Y _i " to the probability of occurrence of Y _i ; the calculation formula of the transaction lift degree Lift (X _i , Y _i ) is:

In the formula, P(Y _i ) represents the probability that the meat duck health representation information is Y _i .

_The degree _{of lift reflects} the _{correlation between} When , the smaller the lift value, the weaker the negative correlation. When Lift(X _i ,Y _i )=1, there is no correlation.

Calculate support and confidence for the environmental monitoring and meat duck health characterization transaction database, set the minimum support threshold and minimum confidence threshold, and search and iterate layer by layer to find strong association rules between multi-source information (while satisfying the minimum Support threshold and minimum confidence threshold rules) For example: "[The temperature inside the house is suitable, the humidity inside the house is suitable, the CO2 concentration is high...] → [Normal behavior, abnormal movement, normal voice, normal body temperature]", Key indicators (such as temperature inside the house, humidity inside the house, CO2 concentration...) are extracted from it and used as important indicators for the establishment of a health and comfort evaluation system for meat ducks.

In the embodiment of the present invention, step S3 includes the following sub-steps:

S31: Use the environmental monitoring information inside and outside the duck house and the meat duck health characterization information as the element layer, use the key indicators of meat duck health and comfort as the indicator layer, and calculate the subjective weight based on the element layer and the indicator layer;

S32: Calculate the objective weight based on the element layer and indicator layer;

S33: Calculate comprehensive subjective and objective weights based on subjective weights and objective weights;

S34: Calculate the meat duck health and comfort index based on the comprehensive subjective weight.

The present invention designs a meat duck health and comfort evaluation system. This evaluation system is applied in step S3. Based on the results of the correlation analysis of environmental monitoring and meat duck health characterization information in step S2, the entire evaluation system is divided into three levels, namely the target layer, element layer and indicator layer. The target layer is the meat duck health and comfort index A; the element layer includes the indoor environment monitoring information U ₁ , the external environment monitoring information U ₂ and the meat duck health characterization information U ₃ ; the indicator layer is the environmental monitoring and meat duck monitoring in step S2 The key indicator C extracted after information association analysis, C={C ₁ , C ₂ , C ₃ ...C _c }, c is the number of key indicators. The health and comfort evaluation of meat duck is:

Subjective weight calculation: Using the element layer as the basis for judgment, the judgment matrix of the element layer is obtained, denoted as PU ₁ , PU ₂ , PU ₃ ; using the indicator layer as the basis for judgment, the judgment matrix of the index layer is obtained, denoted as PC ₁ , PC ₂ , PC ₃ ,…, PC _c . The form of the judgment matrix P is as follows:

In the formula, for the judgment matrix PU of the element layer, b _ij represents the importance of element i relative to element j; for the judgment matrix PC of the indicator layer, b _ij represents the importance of index i relative to index j, and the relationship is as follows:

Compare the importance of elements in each layer and calculate the weight of each element:

Pω= _λmaxω

In the formula, λ _max is the largest characteristic root of R, and ω = (ω ₁ , ω ₂ ..., ω _n ) ^T is the weight vector of P. Using the above formula to calculate the weight of the element layer is ω ^U , the weight of the indicator layer is ω ^C , the weight of the indicator layer relative to the target layer is ω ^CA , and the comprehensive subjective weight is The weight expressions are:

In the formula, u is the number of elements in the feature layer, and c is the number of elements in the indicator layer. will combine subjective weights Denote it as μ ¹ .

Objective weight calculation: After normalizing all the data of each key indicator, average statistics are made for each indicator. The calculation process is as follows: Construct a decision matrix for the meat duck health assessment indicators, recorded as: x = (x _ij ) _m*t , m is the number of key indicator types (the key indicator types in the present invention include environmental information inside the house, environmental information outside the house and meat duck health characterization information), x _ij represents the jth key indicator among the t (1≤j ≤m) The parameter value of the i-th indicator of key indicators. The larger the coefficient of variation value, the greater the weight of the i-th indicator between different evaluation orders.

Comprehensive subjective and objective weights: Use the coefficient of variation to correct the subjective weights.

Meat duck health and comfort evaluation: Use the linear weighting method to calculate the comprehensive score F of the target layer, that is, the meat duck health comfort index. The greater the meat duck health comfort index, the lower the possibility of health abnormalities in the meat duck.

In the embodiment of the present invention, in step S31, the calculation formula of subjective weight μ ¹ is:

In the formula, U represents the number of elements in the feature layer, C represents the number of elements in the indicator layer, and ω represents the weight of each element;

In step S32, the key indicators of meat duck health and comfort are normalized, the comprehensive data value of the i-th indicator after normalization is calculated, the coefficient of variation is calculated based on the comprehensive data value of the i-th indicator, and the coefficient of variation is calculated. Objective weight; among them, the comprehensive data value of the i-th indicator The calculation formulas of the coefficient of variation b _i and objective weight μ ² are respectively:

In the formula, t represents the number of key indicators, Z _il represents the l-th data of the i-th key indicator, x _ij represents the parameter value of the i-th indicator belonging to the j-th type of key indicators among the t key indicators, Represents the comprehensive data value of the jth indicator;

In step S33, the calculation formula of the comprehensive subjective and objective weight β _i is:

In the formula, α _i represents the empirical factor of the i-th indicator, is the subjective weight of the i-th indicator,/> is the objective weight of the i-th indicator;

In step S34, the calculation formula of the meat duck health and comfort index is:

In the formula, Z _i represents the i-th key indicator.

In the embodiment of the present invention, in step S5, the specific method for environmental control is: calculating the deviation and deviation change rate between the environmental information in the duck house and the preset environmental information at the next moment. When the deviation or the deviation change rate is greater than the set threshold, At this time, the predicted value of the meat duck health and comfort index is used as a constraint for environmental regulation.

Compare the predicted value of the environmental information obtained in step S4 with the appropriate environmental information to obtain the deviation E and the deviation change rate ΔE, which together serve as input to the environmental regulation model. Taking the temperature monitoring index indoor temperature as an example, E=HH ₀ , In the formula: H is the temperature prediction value; H ₀ is the suitable temperature in the house. Based on fuzzy theory, the deviation E and deviation change rate △E are converted into fuzzy control theory language values. The specific conversion formula is as follows:

In the formula,

The fuzzy set domain of the prediction value deviation of the indoor temperature is set to [-3,3], which can be divided into 7 levels {-3,-2,-1,0,1,2,3}, which is the input value quantification process. The fuzzy language value of the deviation is: {NB=negative large, NM=negative medium, NS=negative small, Z=0 medium, PM=positive small, PM=positive medium, PB=positive large}. In the same way, the rate of change of the deviation of the predicted value of temperature in the house is also divided into three levels of fuzzy language values (negative, medium, positive) according to the above method; the fuzzy value of the output control variable of the fuzzy control model is: {NB=negative large, NM= Negative center, Z=0 center, PM=positive center, PB=positive large}, convert the fuzzy output language value into the fuzzy control quantity of the temperature control equipment. The specific rules of the environmental control model are as follows: When the deviation E of the predicted value of the indoor temperature and its change rate ΔE are large, the deviations should be eliminated as soon as possible; when they are small or zero, priority should be given to maintaining the stability of the predicted value of the indoor temperature.

Finally, according to the fuzzy control quantity of the temperature control equipment output by the environmental control model, the temperature control equipment control rules were qualitatively analyzed using the predicted value of meat duck health and comfort as constraints to formulate environmental control decisions for caged meat duck breeding.

The caged meat duck breeding environment control method is implemented based on the caged meat duck breeding environment control system, including an environmental monitoring module, a meat duck monitoring module, an environmental control module and a core processing module. Among them, the environmental monitoring module uses environmental monitoring sensors to obtain real-time environmental information inside and outside the duck house (temperature, humidity, wind speed, illumination, CO2 concentration, NH3 concentration, H2S concentration, PM2.5 concentration, PM10 concentration, outside temperature, outside humidity, Wind speed outside the house and illumination outside the house). The meat duck monitoring module uses RGB cameras, sound sensors, motion sensors and body temperature sensors to collect video, sound, movement and body temperature information of meat ducks in real time. The environment control module includes environmental control models and environmental control equipment. The core processing module includes the meat duck health and comfort assessment system and prediction model.

Those of ordinary skill in the art will appreciate that the embodiments described here are provided to help readers understand the principles of the present invention, and it should be understood that the scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and these modifications and combinations are still within the protection scope of the present invention.

Claims (4)

1. A method for regulating the environment of caged meat duck breeding based on health and comfort, which is characterized by including the following steps: S1: Obtain environmental monitoring information inside and outside the duck house and meat duck health characterization information; In the step S1, the meat duck health representation information includes abnormal behavior awareness information, abnormal feed intake awareness information, abnormal movement awareness information, abnormal sound awareness information and abnormal body temperature awareness information; The specific method of obtaining abnormal behavior awareness information and abnormal feed intake awareness information is: collecting duck house videos, and using the adaptive threshold method to obtain meat duck images from the duck house videos, and using the meat duck images as the input of the convolutional neural network. Obtain information on abnormal behavior and abnormal feed intake; The specific method of obtaining sound abnormality awareness information is: collecting meat duck sound data, sequentially performing pre-emphasis, framing, windowing, Fourier transform and Mel frequency cepstrum coefficient solution on the meat duck sound data to obtain the sound feature vector, As sound abnormality awareness information; The specific method to obtain motion abnormality awareness information is: collect meat duck motion data, use signal time domain analysis method, signal frequency domain analysis method and signal time-frequency domain analysis method to extract multi-dimensional features of meat duck motion data, obtain the motion feature vector, and The sound feature vector and the motion feature vector are fused to obtain a fused feature vector, which is used as motion abnormality awareness information; The specific method of obtaining the abnormal body temperature awareness information is: collecting the body temperature data of meat ducks, and using the body temperature data outside the set temperature threshold range as the abnormal body temperature awareness information; S2: Determine the key indicators of health and comfort of meat ducks based on environmental monitoring information inside and outside the duck house and meat duck health characterization information; S21: Sequentially pair the environmental monitoring information inside and outside the duck house with the meat duck health characterization information and add time information. Use the collection of environmental monitoring information inside and outside the duck house as the environmental monitoring transaction database, and use the collection of meat duck health characterization information as the meat duck The health characterization transaction database uses the collection of environmental monitoring transaction database, meat duck health characterization transaction database and time information as the general transaction database; S22: Calculate transaction support and transaction confidence based on the total transaction database; S23: Calculate the transaction improvement degree based on the transaction support and transaction confidence; S24: Use data whose transaction improvement degree is greater than the set improvement threshold as a key indicator of the health and comfort of meat ducks; S3: Determine the meat duck health and comfort index based on the key indicators of meat duck health and comfort; The step S3 includes the following sub-steps: S31: Use the environmental monitoring information inside and outside the duck house and the meat duck health characterization information as the element layer, use the key indicators of meat duck health and comfort as the indicator layer, and calculate the subjective weight based on the element layer and the indicator layer; S32: Calculate the objective weight based on the element layer and indicator layer; S33: Calculate comprehensive subjective and objective weights based on subjective weights and objective weights; S34: Calculate the meat duck health and comfort index based on the comprehensive subjective weight; S4: Use the environmental monitoring information inside and outside the duck house and the health and comfort index of meat ducks as the input of the neural network to obtain the predicted value of the environmental information inside the duck house and the health and comfort index of meat ducks at the next moment; S5: Carry out environmental regulation based on the environmental information in the duck house at the next moment and the predicted value of the meat duck health and comfort index. 2. The caged meat duck breeding environment control method based on health and comfort according to claim 1, characterized in that, in the step S22, the calculation formula of the transaction support Support ( Xi _, Yi ₎ is: _In the formula , P _{( _} _ _{_} _ _ _{_} _ _{_} Represents the number of transactions containing environmental monitoring information inside and outside the duck house in the X _i state and meat duck health characterization information in the Y _i state, number ( ALLSamples ) represents the number of transactions in the total transaction library; In the step S22, the calculation formula of the transaction confidence Confidence ( X _i → Y _i ) is: In _the formula, P ( Xi Y _i ) represents the ratio of transactions _containing meat duck health characterization information in the general transaction database with environmental monitoring information inside and outside the duck house in the Xi state . P ( Xi ₎ represents the ratio of the environmental monitoring information inside and outside the duck house in _the Xi state _. The probability; In the step S23, the calculation formula of the transaction lifting degree Lift ( X _i , Y _i ) is: In the formula, P ( Y _i ) represents the probability that the meat duck health representation information is in the Y _i state. 3. The caged meat duck breeding environment regulation method based on health and comfort according to claim 1, characterized in that, in the step S31, the calculation formula of the subjective weight μ ¹ is: In the formula, U represents the number of elements in the feature layer, C represents the number of elements in the indicator layer, and ω represents the weight of each element; In the step S32, the key indicators of health and comfort of meat ducks are normalized, the comprehensive data value of the i-th indicator after normalization is calculated, and the variation coefficient is calculated based on the comprehensive data value of the i- th indicator. The coefficient calculates the objective weight; among them, the comprehensive data value of the i- th indicator , the calculation formulas of variation coefficient b _i and objective weight μ ² are respectively: In the formula, t represents the number of key indicators, Z _il represents the l- th data of the i- th key indicator, x _ij represents the parameter value of the i -th indicator belonging to the j-th type of key indicators among the t key indicators, Represents the comprehensive data value of the jth indicator; In the step S33, the calculation formula of the comprehensive subjective and objective weight β _i is: In the formula, α _i represents the empirical factor of the i -th indicator, is the subjective weight of the i -th indicator,/> is the objective weight of the i -th indicator; In the step S34, the calculation formula of the meat duck health and comfort index is: In the formula, Z _i represents the i -th key indicator. 4. The caged meat duck breeding environment control method based on health and comfort according to claim 1, characterized in that in step S5, the specific method for environmental control is: calculating the environmental information in the duck house at the next moment. The deviation and deviation change rate from the preset environmental information. When the deviation or deviation change rate is greater than the set threshold, the predicted value of the meat duck health and comfort index is used as a constraint for environmental regulation.