A method and system for detecting the health status of broiler chickens based on Markov chain

Abstract

The present invention provides a method for detecting the health status of broiler chickens based on Markov chains, which relates to the technical field of intelligent breeding and includes: obtaining the behavior data of the target broiler chickens; cleaning, standardizing, feature extracting and feature synthesizing the behavior data to obtain time series detection data for detection; dividing the health status of the broiler chickens according to the preset historical data set of broiler chickens to obtain multiple health statuses; constructing a Markov chain model according to the health statuses; training the Markov chain model by using the preset historical data set of the health status of broiler chickens to optimize the initial probability of each health status and the state transition probability matrix, so as to obtain a trained detection model for the health status of broiler chickens; and inputting the time series detection data into the detection model for the health status of broiler chickens to obtain a detection result. The present invention can combine multi-dimensional data, realize real-time monitoring and intelligent decision support, and improve the efficiency and level of broiler chicken breeding management.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent breeding, and particularly to a method and system for detecting the health status of broiler chickens based on Markov chains. Background Art

[0002] The health status of broiler chickens directly affects their growth rate and final meat quality. Timely and accurately detecting the health status of broiler chickens is of great significance to the broiler chicken breeding industry.

[0003] The invention patent with the application number 201310547402.9 discloses a method for real-time monitoring of the physical health of poultry based on data extraction, including the following steps: A1: constructing a regional monitoring wireless sensor network, respectively collecting the characterization data of healthy and diseased poultry within 720 hours, using data extraction technology for data processing, extracting feature information, and constructing a health pattern library and a disease pattern library for the physical condition of poultry; A2: real-time monitoring of the physical health of poultry, real-time collecting the relevant characterization data of poultry, extracting features through data extraction, respectively matching with the health pattern library and the disease pattern library, updating the corresponding pattern library according to the matching results, and sending a report to the administrator to take corresponding measures. This technology combines data extraction and wireless sensor networks to real-time monitor the physical condition of poultry, is applicable to large and medium-sized breeding farms, and can initially judge the health status of poultry through the detected poultry data, so as to detect diseases in time and prevent and control them quickly, so as to reduce or even avoid the economic losses caused by the spread of poultry diseases.

[0004] Although the method for real-time monitoring of the physical health of poultry based on data extraction of this technology has certain advantages in real-time monitoring and feature extraction, it does not consider the evolution process of the health status of broiler chickens, resulting in relatively insufficient modeling and prediction capabilities for the health status of poultry, and restricting the comprehensiveness and accuracy of the monitoring system. Summary of the Invention

[0005] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a method and system for detecting the health status of broiler chickens based on Markov chains.

[0006] To achieve the above purpose, the present invention provides the following solutions:

[0007] A method for detecting the health status of broiler chickens based on Markov chains includes:

[0008] Obtaining the behavior data of the target broiler chicken; the behavior data includes: activity frequency, food intake, water intake, and call frequency;

[0009] Cleaning, standardizing, feature extracting, and feature synthesizing the behavior data to obtain time series detection data for detection;

[0010] Divide the health status of broiler chickens according to a preset historical dataset of broiler chickens to obtain multiple health statuses;

[0011] Construct a Markov chain model based on the health status;

[0012] Train the Markov chain model using a preset historical dataset of broiler chicken health status to optimize the initial probability of each health status and the state transition probability matrix, and obtain a trained broiler chicken health status detection model;

[0013] Input the time series detection data into the broiler chicken health status detection model to obtain a detection result.

[0014] Preferably, the health statuses include: healthy, sub-healthy, mild disease, and severe disease.

[0015] Preferably, obtain the behavior data of the target broiler chicken, including:

[0016] Set an infrared sensor in the activity area of the target broiler chicken, and use the infrared sensor to obtain the activity frequency of the target broiler chicken;

[0017] Respectively set a weight sensor and a flow sensor on the feeder and waterer of the target broiler chicken, and respectively obtain the food intake and water intake according to the weight sensor and the flow sensor;

[0018] Set a sound sensor in the activity area of the target broiler chicken, and collect the call frequency according to the sound sensor.

[0019] Preferably, clean, standardize, and extract features from the behavior data to obtain a health feature dataset for detection, including:

[0020] Perform outlier processing, missing value filling, and data standardization on each type of behavior data in sequence to obtain standardized data;

[0021] Extract features from the standardized data to obtain activity frequency features, food intake features, water intake features, and call frequency features;

[0022] Integrate the activity frequency features, food intake features, water intake features, and call frequency features to form multiple health feature datasets for detection;

[0023] Sample the health feature dataset according to a preset time window to form the time series detection data.

[0024] Preferably, the activity frequency characteristics include the number of activities, activity intensity, activity time, and rest time within a unit time; the food intake characteristics include the number of food intakes, the amount of food intake per time, and the total daily food intake; the water intake characteristics include the number of water intakes, the amount of water intake per time, and the total daily water intake; the call frequency characteristics include the call frequency within a unit time, the duration of each call, and the call volume.

[0025] Preferably, a Markov chain model is constructed according to the health status, including:

[0026] Define the health status of broiler chickens as a discrete finite state set S = {s 1 , s 2 , …, s n}, where s i represents the i-th health status;

[0027] Assume that the probability distribution of the broiler chicken state at the initial time t = 0 is π = (π 1 , π 2 , …, π n ); where π i = P(S 0 = s i ); S 0 is the system state at the initial time t = 0, and P(S 0 = s i ) represents the probability that the system is in the state s i at the initial time;

[0028] Construct a state transition probability matrix P; where P ij = P(S t+1 = s j | S t = s i ), and t is the time step;

[0029] Construct the Markov chain model according to the finite state set, the probability distribution, and the state transition probability matrix.

[0030] Preferably, the historical data set of broiler chicken health status includes: preset activity frequency characteristic data, food intake characteristic data, water intake characteristic data, and call frequency characteristic data.

[0031] According to the specific embodiments provided by the present invention, the following technical effects of the present invention are disclosed:

[0032] The present invention provides a method for detecting the health status of broiler chickens based on Markov chains, including: obtaining the behavioral data of the target broiler chicken; the behavioral data includes: activity frequency, food intake, water intake, and call frequency; cleaning, standardizing, feature extracting, and feature synthesizing the behavioral data to obtain time series detection data for detection; dividing the health status of broiler chickens according to a preset historical dataset of broiler chickens to obtain multiple health statuses; constructing a Markov chain model according to the health statuses; training the Markov chain model using a preset historical dataset of the health status of broiler chickens to optimize the initial probability and state transition probability matrix of each health status, and obtaining a trained detection model for the health status of broiler chickens; inputting the time series detection data into the detection model for the health status of broiler chickens to obtain a detection result. The method for detecting the health status of broiler chickens based on Markov chains according to the present invention can combine multi-dimensional data, achieve real-time monitoring and intelligent decision support, and improve the efficiency and level of broiler chicken breeding management. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0034] Figure 1 It is a flowchart of the method provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0036] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0037] Figure 1 It is a flowchart of the method provided by the embodiment of the present invention. As Figure 1 shown, the present invention provides a method for detecting the health status of broiler chickens based on Markov chains, including:

[0038] Step 100: Obtain the behavioral data of the target broiler chicken; the behavioral data includes: activity frequency, food intake, water intake, and call frequency;

[0039] Step 200: Clean, standardize, extract features and synthesize features from the behavior data to obtain time series detection data for detection;

[0040] Step 300: Divide the health status of the broiler chickens according to the preset historical dataset of broiler chickens to obtain multiple health statuses;

[0041] Step 400: Construct a Markov chain model according to the health status;

[0042] Step 500: Use the preset historical dataset of broiler chicken health status to train the Markov chain model to optimize the initial probability and state transition probability matrix of each health status, and obtain a trained broiler chicken health status detection model;

[0043] Step 600: Input the time series detection data into the broiler chicken health status detection model to obtain a detection result.

[0044] Preferably, the health statuses include: healthy, sub-healthy, mild disease and severe disease.

[0045] Preferably, obtaining the behavior data of the target broiler chickens includes:

[0046] Set an infrared sensor in the activity area of the target broiler chicken, and use the infrared sensor to obtain the activity frequency of the target broiler chicken;

[0047] Respectively set a weight sensor and a flow sensor on the feeder and waterer of the target broiler chicken, and respectively obtain the food intake and water intake according to the weight sensor and the flow sensor;

[0048] Set a sound sensor in the activity area of the target broiler chicken, and collect the call frequency according to the sound sensor.

[0049] Specifically, in this embodiment, the installation of the equipment and the collection of data are first carried out as follows:

[0050] Step 101: Equipment selection and installation. Install an activity monitoring sensor (such as an infrared sensor) around the cage. Install a weight sensor under the trough of the feeder to record the reduction of food in real time. Install a flow sensor at the water outlet and water inlet of the waterer to monitor the water intake. Install a sound sensor or microphone to capture the call frequency and intensity of the broiler chickens. Reasonably install the above sensor devices in the cages, feeders, waterers and activity areas of the breeding farm to ensure that the data of each target broiler chicken can be accurately collected.

[0051] Step 102, Data collection and transmission: Record the number of movements and amplitudes of broiler chickens within a certain period of time to detect their movement behaviors; Measure the weight change of the feed in the feeder in real time and record the amount of each feeding; Measure the water volume change in the waterer in real time and record the amount of each drinking; Collect the crowing data of broiler chickens through a sound sensor, including frequency, duration, and intensity.

[0052] Step 103, Data transmission: Use a wireless sensor network (such as WiFi, ZigBee) to transmit the collected behavior data to the central data processing system in real time. Set up a data receiving point (such as a data gateway) to centrally collect sensor data and transmit it to the server for further processing.

[0053] Preferably, clean, standardize, and extract features from the behavior data to obtain a healthy feature dataset for detection, including:

[0054] Perform outlier processing, missing value filling, and data standardization on various types of the behavior data in sequence to obtain standardized data;

[0055] Extract features from the standardized data to obtain activity frequency features, food intake features, water intake features, and crowing frequency features;

[0056] Integrate the activity frequency features, the food intake features, the water intake features, and the crowing frequency features to form multiple healthy feature datasets for detection;

[0057] Sample the healthy feature dataset according to a preset time window to form the time series detection data.

[0058] Furthermore, the data preprocessing steps of this embodiment include:

[0059] Step 201, Data cleaning, including: Outlier processing, identifying and removing outliers in the data (such as food intake or water intake data that significantly exceed the normal range); Missing value filling, using appropriate methods (such as mean filling, interpolation method) to fill the missing values in the collected data.

[0060] Step 202, Data standardization: Normalize the behavior data recorded by different sensors to make the data scales consistent, suitable for subsequent feature extraction and model calculation.

[0061] Even further, the feature extraction steps of this embodiment include:

[0062] Step 203, Calculate the number of activities, activity intensity, activity time, stationary time, etc. per unit time.

[0063] Step 204, Calculate the number of feeding times, the amount of each feeding, the total daily food intake, etc. per unit time.

[0064] Step 205, calculate the number of drinking times, the amount of water drunk each time, the total daily water intake, etc. within a unit time.

[0065] Step 206, calculate the call frequency, the duration of each call, the call volume, etc. within a unit time.

[0066] Preferably, the activity frequency characteristics include the number of activities, activity intensity, activity time, and stationary time within a unit time; the food intake characteristics include the number of feeding times, the amount of food eaten each time, and the total daily food intake; the water intake characteristics include the number of drinking times, the amount of water drunk each time, and the total daily water intake; the call frequency characteristics include: the call frequency, the duration of each call, and the call volume within a unit time.

[0067] Furthermore, this embodiment further includes the steps of generating a health feature dataset, specifically:

[0068] Step 207, integrate the feature data collected and processed by each sensor to form multiple health feature datasets for detection, corresponding to each broiler respectively.

[0069] Step 208, sample the feature data according to a certain time window (such as hours, days) to form time series data.

[0070] Preferably, construct a Markov chain model according to the health status, including:

[0071] Define the health status of the broiler as a discrete finite state set S = {s 1 , s 2 , …, s n}, where s i represents the i-th health status;

[0072] Assume that the probability distribution of the broiler state at the initial time t = 0 is π = (π 1 , π 2 , …, π n ); where π i = P(S 0 = s i ); S 0 is the system state at the initial time t = 0, and P(S 0 = s i ) represents the probability that the system is in the state s i at the initial time;

[0073] Construct a state transition probability matrix P; where, P ij = P(S t+1 = s j|S t = s i ), where t is the time step;

[0074] Construct the Markov chain model according to the finite state set, the probability distribution, and the state transition probability matrix.

[0075] Specifically, assume that the health status of broilers is divided into four discrete states S = {healthy, sub - healthy, mildly ill, severely ill}, and s 1 , s 2 , s 3 , s 4 are used to represent these four states respectively. Then P(S 0 = s 1 represents the probability that a certain broiler is in a healthy state at the beginning of the monitoring. The sum of the probability distributions of all initial states should be 1, that is where n is the total number of states. In the example of this embodiment: P(S 0 = s 1 ) + P(S 0 = s 2 ) + P(S 0 = s 3 ) + P(S 0 = s 4 ) = 1.

[0076] Furthermore, the state probability P(S 0 = s i ) at the initial moment can usually be statistically analyzed based on historical data. If there is a large amount of historical data, we can estimate these probabilities by counting the frequencies of various states at the beginning of the monitoring in the past. For example:

[0077] In the past 1000 monitoring cycles, there were 600 times when the broiler was in a healthy state at the beginning, then P(S 0 = s 1 ) ≈ 0.6.

[0078] There were 200 times when the broiler was in a sub - healthy state at the beginning, then P(S 0 = s 2 ) ≈ 0.2.

[0079] There were 150 times when the broiler was in a mildly ill state at the beginning, then P(S 0 = s 3 ) ≈ 0.15.

[0080] There were 50 times when the broiler was in a severely ill state at the beginning, then P(S 0 = s 4 ) ≈ 0.05.

[0081] Preferably, the historical dataset of the broiler health status includes: preset activity frequency feature data, food intake feature data, water intake feature data, and call frequency feature data.

[0082] Specifically, the Markov chain model of this embodiment further includes an observation matrix O, which describes the probability of observing specific observation values (sensor data) in each state.

[0083] The beneficial effects of the present invention are as follows:

[0084] (1) The present invention utilizes the behavioral data of broilers, including multi-dimensional information such as activity frequency, food intake, water intake, and call frequency. By comprehensively analyzing these data, a more comprehensive understanding of the broiler health status can be achieved.

[0085] (2) The present invention obtains time series data for detection by cleaning, standardizing, feature extracting, and feature synthesizing the behavioral data. This enables dynamic monitoring and analysis of the broiler health status, thus timely detecting status changes.

[0086] (3) Based on the preset historical dataset of broilers, the present invention divides the broiler health status and constructs a Markov chain model. Through the modeling and training of historical data, more accurate inference and prediction of the broiler health status can be made.

[0087] (4) The Markov chain model of the present invention can effectively compress and represent the complexity of state transitions, enabling the relationship and transition rules between broiler health states to be described in a concise manner.

[0088] (5) After obtaining new time series detection data, the trained broiler health status detection model of the present invention can detect and diagnose the broiler health status in real time, helping breeders take corresponding measures in a timely manner.

[0089] (6) The present invention can provide intelligent decision support. Through model analysis and detection results, it helps breeders formulate more reasonable breeding management strategies, improving the broiler health level and production efficiency.

[0090] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0091] Specific examples are used in this article to elaborate on the principles and implementation methods of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A broiler health status detection method based on Markov chain, characterized in that: include: Obtain behavioral data of target broilers; The behavioral data include: activity frequency, food intake, water intake and call frequency; Cleaning, standardizing, feature extracting and feature synthesizing the behavior data to obtain time series detection data for detection; The health status of broilers is divided according to a preset broiler historical data set to obtain multiple health statuses; constructing a Markov chain model according to the health status; The Markov chain model is trained using a preset broiler health status historical data set to optimize the initial probability and state transition probability matrix of each health state, thereby obtaining a trained broiler health status detection model; Inputting the time series detection data into a broiler health status detection model to obtain a detection result; A Markov chain model is constructed according to the health state, including: The health status of broiler chickens is defined as a discrete finite state set S = {s1, s2, ..., s n }, where s i represents the i-th health state; Assume that the probability distribution of the broiler chicken state at the initial time t = 0 is π = (π1, π2, ..., π n );where π i =P(S0=s i ); S0 is the system state at the initial time t = 0, P(S0 = s i ) indicates that the system is in state s at the initial moment i probability; Construct state transition probability matrix P; Among them, P ij =P(S t+1 =s j ∣S t =s i ), t is the time step; The Markov chain model is constructed according to the finite state set, the probability distribution and the state transition probability matrix.

2. The broiler health status detection method based on Markov chain according to claim 1, characterized in that: The health status includes: healthy, sub-healthy, mild disease and severe disease.

3. The broiler health status detection method based on Markov chain according to claim 1, characterized in that: Obtain behavioral data of target broilers, including: Setting an infrared sensor in the activity area of ​​the target broiler chicken, and using the infrared sensor to obtain the activity frequency of the target broiler chicken; A weight sensor and a flow sensor are respectively arranged on the feeder and the drinker of the target broiler, and the food intake and the water intake are respectively obtained according to the weight sensor and the flow sensor; A sound sensor is arranged in the activity area of ​​the target broiler, and the calling frequency is collected according to the sound sensor.

4. The broiler health status detection method based on Markov chain according to claim 1, characterized in that: The behavior data is cleaned, standardized and feature extracted to obtain a health feature data set for detection, including: Performing outlier processing, missing value filling and data standardization on each type of behavioral data in turn to obtain standardized data; Extracting features from the standardized data to obtain activity frequency features, food intake features, water intake features, and call frequency features; Integrating the activity frequency feature, the food intake feature, the water intake feature and the call frequency feature to form a plurality of health feature data sets for detection; The health feature data set is sampled according to a preset time window to form the time series detection data.

5. The broiler health status detection method based on Markov chain according to claim 1, characterized in that: The activity frequency characteristics include the number of activities per unit time, activity intensity, activity time and static time; the food intake characteristics include the number of meals per unit time, the amount of food eaten each time and the total amount of food eaten per day; The water drinking characteristics include the number of times of drinking per unit time, the amount of water drunk each time and the total amount of water drunk per day; The call frequency characteristics include: the call frequency per unit time, the duration of each call and the call volume.

6. The broiler health status detection method based on Markov chain according to claim 1, characterized in that: The broiler health status historical data set includes: preset activity frequency characteristic data, food intake characteristic data, water intake characteristic data and call frequency characteristic data.

Images