A method and system for rapidly predicting social hierarchy of domestic chickens based on early warning behavior

Abstract

The application relates to the technical field of poultry house biological monitoring, and particularly relates to a domestic chicken social grade rapid prediction method and system based on early warning behaviors. The method comprises the following steps: performing a standardized early warning behavior test on a domestic chicken group in a target area, and collecting reaction characteristics; integrating comprehensive behavior characteristic vectors of each domestic chicken individual; inputting the comprehensive behavior characteristic vectors of each domestic chicken individual into a preset social grade prediction model as training samples, and training the social grade prediction model by using a pre-constructed verification set to obtain a trained social grade prediction model; obtaining a fitting coefficient; determining a predicted grade score of the domestic chicken individual according to the comprehensive behavior characteristic vectors of the domestic chicken individual and the fitting coefficient, and determining a predicted social grade of the domestic chicken individual according to the predicted grade score. The application can accurately predict the social grade of the domestic chicken individual, and solves the problems of long cycle, strong subjectivity, great interference and difficulty in standardization in the prior art.

Description

Technical Field

[0001] This application relates to the field of poultry house biological monitoring technology, and in particular to a method and system for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior. Background Technology

[0002] Social hierarchy refers to the priority order of animals in areas such as feeding, resting, territoriality, and mating, and exists in all social animals. Domestic chickens were the first social animal species found to exhibit social hierarchy. Social hierarchy directly affects an individual's resource acquisition, aggressive behavior, and stress exposure levels; therefore, studying the impact of social hierarchy on livestock and poultry production performance and welfare is crucial. High-ranking individuals have priority in feeding, use of laying boxes, and sand baths; low-ranking individuals are forced to eat at off-peak times, have insufficient feed intake, and are in a state of chronic social stress, leading to overactivation of the HPA axis, immunosuppression, and behavioral abnormalities.

[0003] Currently, the main methods for assessing the social hierarchy of chicken flocks include the following categories: 1. Behavioral observation method: This method involves manually observing and recording interactions between individuals, such as aggression and submission, to construct a hierarchy matrix. This method requires continuous observation for several weeks, is time-consuming and labor-intensive, and is highly subjective.

[0004] 2. Competition Test Method: This method determines the rank of chickens by observing the competition results of individuals within a limited resource environment (such as a single feeding point). This method significantly disrupts normal flock production and is heavily dependent on environmental conditions.

[0005] 3. Data analysis method: Based on historical production data (such as feed intake, weight changes, etc.), the level is inferred by retrospection, but it has the drawbacks of difficulty in data acquisition and lagging evaluation.

[0006] 4. Sensor monitoring method: In recent years, some studies have attempted to use inertial sensors to obtain chicken activity data to identify flock order, but each chicken needs to be equipped with a device, which is difficult to promote in large-scale farming scenarios, and still requires a long data collection cycle.

[0007] In summary, existing social rating methods generally suffer from problems such as long cycles, strong subjectivity, large interference, and difficulty in standardization, and there is an urgent need for a new, fast, objective, and standardized evaluation method. Summary of the Invention

[0008] To address the aforementioned issues, this application provides a method and system for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior, which can solve at least one of the problems in the prior art.

[0009] According to one aspect of the embodiments of this application, a method for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior is proposed, the method comprising: During the model training phase, a standardized early warning behavior experiment is conducted on a group of domestic chickens in the target area. The reaction characteristics of each individual domestic chicken in the group are collected within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area within the target area, the intensity of movement, and the frequency of risk avoidance behavior. For each individual chicken, the chicken's reaction latency, the proportion of time spent in the resource area, the intensity of movement, and the frequency of risk avoidance behavior within the preset time period are integrated into a comprehensive behavioral feature vector for the chicken. The comprehensive behavioral feature vectors of each of the domestic chicken individuals are used as training samples and input into the preset social ranking prediction model. The social ranking prediction model is trained using a pre-constructed validation set to obtain the trained social ranking prediction model. Obtain the fitting coefficients for the response features output by the trained social rank prediction model; During the inference phase, for each individual chicken, the predicted level score of the chicken is determined based on the comprehensive behavioral feature vector of the chicken and the fitting coefficient output by the trained social level prediction model, and the predicted social level of the chicken is determined based on the predicted level score of the chicken.

[0010] In the above scheme, the standardized early warning behavior experiment on domestic chickens in the target area includes: A standardized sound stimulus consisting of a preset sound intensity, a preset sound frequency, and a preset sound duration is applied to the chicken population within the target area to provide sound stimulation to each chicken population.

[0011] In the above scheme, the reaction latency is the time interval from the emission of the standardized sound stimulus to the individual chicken taking action; the resource area is located at the center of the target area; the dwell time ratio is the proportion of the dwell time of the individual chicken in the resource area to the preset time; the movement intensity is used to characterize the distance the individual chicken moves within the preset time; and the risk avoidance behavior frequency is the number of times the individual chicken moves away from the resource area or the center of the chicken group within the preset time.

[0012] In the above scheme, the step of inputting the comprehensive behavioral feature vector of each of the chicken individuals as a training sample into a preset social ranking prediction model, and training the social ranking prediction model using a pre-constructed validation set to obtain the trained social ranking prediction model includes: For each individual chicken, the comprehensive behavioral feature vector of the individual chicken is input into the social rank prediction model to obtain the predicted rank score corresponding to the individual chicken output by the social rank prediction model. The predicted rank score is calculated by the fitting coefficient currently output by the social rank prediction model and the preset fitting formula. The predicted social rank of the individual chicken is determined based on the predicted rank score. In the validation set, the true social class of the chicken individual is determined, and the loss value corresponding to the chicken individual is determined based on the quantitative difference between the predicted social class and the true social class of the chicken individual. The social ranking prediction model is trained based on the loss value corresponding to each individual chicken to iteratively optimize the fitting coefficients and obtain the trained social ranking prediction model.

[0013] In the above scheme, the preset fitting formula is expressed by the following formula: Rank i = β0 + β1·L i norm + β2·P i + β3·I i + β4·F i In the formula, Rank i The predicted grade score for the i-th domestic chicken individual, β0, β1, β2, β3, and β4 constitute the fitting coefficients, L i norm Let P be the latency period of the reaction of the i-th domestic chicken individual. i I represents the proportion of time the i-th domestic chicken individual spends in the resource area. i Let F be the exercise intensity of the i-th domestic chicken individual. i Let be the frequency of the risk avoidance behavior of the i-th domestic chicken individual.

[0014] In the above scheme, during the inference phase, for each individual chicken, the predicted grade score of the chicken is determined based on the comprehensive behavioral feature vector of the chicken and the fitting coefficient output by the trained social grade prediction model, including: During the inference phase, for each individual chicken, the response latency, the proportion of time spent in the resource area within the target region, the movement intensity, and the frequency of risk avoidance behavior in the comprehensive behavioral feature vector of the individual chicken, as well as the fitting coefficients output by the trained social rank prediction model, are input into the fitting formula to obtain the predicted rank score of the individual chicken.

[0015] In the above scheme, determining the predicted social class of the chicken individual based on its predicted class score includes: If the predicted social class score of the individual chicken is within the first preset score range, then the predicted social class of the individual chicken is determined to be high. If the predicted social grade of the individual chicken is within the second preset score range, then the predicted social grade of the individual chicken is determined to be intermediate. If the predicted social grade score of the individual chicken is within the third preset score range, then the predicted social grade of the individual chicken is determined to be low. Wherein, the maximum value in the first preset scoring range is less than the minimum value in the second preset scoring range, and the maximum value in the second preset scoring range is less than the minimum value in the third preset scoring range.

[0016] In the above scheme, after determining the predicted social class of the chicken individual based on its predicted class score, the method further includes: For each individual chicken, multiple physiological indicators were determined, including serum corticosterone, inflammatory factors, and gene expression in key brain regions. Based on the serum corticosterone levels of each individual chicken and the predicted social class of each individual chicken, the biological verification of serum corticosterone levels among different social classes of the chicken population was performed. Based on the inflammatory factors of each individual chicken and the predicted social class of each individual chicken, the biological verification of inflammatory factors between different social classes of the chicken population was carried out. Biological verification of key brain region gene expression across different social ranks of the chicken population was conducted based on the gene expression of key brain regions in each individual chicken and the predicted social rank of each individual chicken.

[0017] According to one aspect of the embodiments of this application, a rapid prediction system for the social hierarchy of domestic chickens based on early warning behavior is also proposed, the system comprising: The data acquisition unit is used to conduct a standardized early warning behavior experiment on a group of chickens in the target area during the model training phase, and to collect the reaction characteristics of each individual chicken in the chicken group within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area within the target area, the movement intensity, and the frequency of risk avoidance behavior. The integration unit is used to integrate the response latency, the proportion of time spent in the resource area, the movement intensity, and the frequency of risk avoidance behavior of each chicken individual within a preset time period into a comprehensive behavioral feature vector of the chicken individual. The training unit is used to input the comprehensive behavioral feature vector of each of the chicken individuals as training samples into the preset social ranking prediction model, and to train the social ranking prediction model using a pre-built validation set to obtain the trained social ranking prediction model. An acquisition unit is used to acquire the fitting coefficients for the response features output by the trained social rank prediction model. The prediction unit is used in the inference phase to determine the predicted level score of each chicken individual based on the comprehensive behavioral feature vector of the chicken individual and the fitting coefficient output by the trained social level prediction model, and to determine the predicted social level of the chicken individual based on the predicted level score.

[0018] The beneficial effects of this application are as follows: This application uses the comprehensive behavioral feature vector of each individual chicken as a training sample and inputs it into a preset social rank prediction model. It then uses a pre-constructed validation set to train the social rank prediction model, iteratively optimizing the fitting coefficients to obtain the fitting coefficients output by the trained social rank prediction model. These coefficients are then used to calculate the predicted rank score for each individual chicken, thus obtaining its predicted social rank. This application can train the social rank prediction model by comprehensively considering four response features: latency period, the proportion of time spent in the resource area within the target region, movement intensity, and the frequency of avoidance behavior. This provides a comprehensive behavioral feature vector of the chicken after being stimulated, from multiple dimensions, making the predicted rank score obtained by the trained social rank prediction model closer to the true value, thus more accurately predicting the social rank of the chicken. This solves the problems of existing methods relying on manual behavioral observation and competitive testing, such as long cycles, strong subjectivity, large interference, and difficulty in standardization. Attached Figure Description

[0019] Figure 1 A flowchart illustrating the rapid prediction method for the social hierarchy of domestic chickens based on early warning behavior, provided in this application embodiment; Figure 2 This is an overall system architecture diagram provided in the embodiments of this application; Figure 3 This is an overall flowchart provided in the embodiments of this application; Figure 4 This is a schematic diagram of biological verification provided in an embodiment of this application; Figure 5 This is another schematic diagram of biological verification provided in the embodiments of this application; Figure 6 This is a schematic diagram of the target area provided in the embodiments of this application; Figure 7This is a schematic diagram of the internal training of the model provided in the embodiments of this application; Figure 8 This is a schematic diagram of the dual verification provided in the embodiments of this application; Figure 9 This is a complete schematic diagram of the model reasoning process used in the embodiments of this application; Figure 10 This is a block diagram of a rapid prediction system for the social hierarchy of domestic chickens based on early warning behavior, provided in an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the solutions of this application, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] It should be noted that while some processes described in the specification, claims, and accompanying drawings include multiple steps appearing in a specific order, it should be clearly understood that these steps may not be performed in the order they appear herein, or may be performed in parallel. The step numbers are merely used to distinguish different steps and do not themselves represent any execution order. Furthermore, descriptions such as "first," "second," or "objective" in this document are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. "Multiple" in this document refers to at least two.

[0022] It is worth noting that, in the specific embodiments of this application, relevant data such as reaction characteristics (including reaction latency, the proportion of time spent in resource areas within the target area, movement intensity, and frequency of risk avoidance behavior) and comprehensive behavioral feature vectors are involved. When the above embodiments of this application are applied to specific products or technologies, permission or consent from the target object is required, and the collection, use, and processing of relevant data must comply with relevant laws, regulations, and standards. For example, when this application embodiment needs to obtain relevant data such as reaction characteristics (including reaction latency, the proportion of time spent in resource areas within the target area, movement intensity, and frequency of risk avoidance behavior) and comprehensive behavioral feature vectors, separate permission or consent from the target object can be obtained through pop-up windows or redirection to a confirmation page. After obtaining separate permission or consent from the target object, the relevant data such as reaction characteristics (including reaction latency, the proportion of time spent in resource areas within the target area, movement intensity, and frequency of risk avoidance behavior) and comprehensive behavioral feature vectors used to enable the embodiments of this application to operate normally are then obtained.

[0023] The following provides a detailed description of the specific implementation methods of the embodiments of this application: Please see Figure 1 , Figure 1 This is a flowchart illustrating the rapid prediction method for the social status of domestic chickens based on early warning behavior provided in the embodiments of this application. Figure 1 The method for rapid prediction of social hierarchy in domestic chickens based on early warning behavior, as shown, includes: Step 210: During the model training phase, a standardized early warning behavior experiment is conducted on the chicken population in the target area. The reaction characteristics of each individual chicken in the chicken population are collected within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area in the target area, the movement intensity, and the frequency of risk avoidance behavior. Step 220: For each individual chicken, integrate the chicken's reaction latency, the proportion of time spent in the resource area, the intensity of movement, and the frequency of risk avoidance behavior within the preset time period into a comprehensive behavioral feature vector for the chicken. Step 230: Input the comprehensive behavioral feature vector of each of the chicken individuals as a training sample into the preset social ranking prediction model, and train the social ranking prediction model using the pre-built validation set to obtain the trained social ranking prediction model. Step 240: Obtain the fitting coefficients for the response features output by the trained social rank prediction model; Step 250: In the inference phase, for each individual chicken, the predicted level score of the individual chicken is determined based on the comprehensive behavioral feature vector of the individual chicken and the fitting coefficient output by the trained social level prediction model, and the predicted social level of the individual chicken is determined based on the predicted level score of the individual chicken.

[0024] First, the overall system architecture of this application can be as follows: Figure 2 As shown, the overall system architecture of this application includes a standardized sound stimulus module for implementing standardized early warning behavior experiments, a video acquisition module for collecting response features, a behavior recognition and analysis module for distinguishing the response features of individual chickens, a feature extraction module for generating comprehensive behavior feature vectors, a social level prediction module for outputting prediction level scores, and a biological verification module for conducting biological verification.

[0025] The following combination Figure 3 The embodiments of this application will be described in detail below: Step 1: Social Rank Gold Standard Assessment – ​​Select chickens of similar age with similar physical condition, group them, and establish a social rank ranking through feeding competition and relative position competition (build a validation set, which includes the true social rank of each individual chicken) as a reference standard for model training.

[0026] Step 2: Standardized warning behavior experiment - Apply standardized sound stimulation to a group of chickens in a quiet environment, and use pre-installed camera equipment to record individual behavioral responses within a preset time (e.g., 5 minutes) after the stimulation to obtain response characteristics.

[0027] Step 3: Response Feature Extraction and Quantification – Use video analysis software to automatically track the video recordings and quantify and extract the warning behavior feature parameters (response features) of each individual chicken.

[0028] Step 4: Construction of the social rank prediction model – Using the social rank obtained in Step 1 as the dependent variable and the response features extracted in Step 3 (i.e., the integrated comprehensive behavioral feature vector) as the independent variable, a social rank prediction model is established using multivariate statistical or machine learning methods.

[0029] Step 5: Model Validation (e.g.) Figure 8 (As shown) — ① Accuracy verification: Compare the consistency between the predicted social rank obtained by the social rank prediction model and the rank assessment by traditional methods (such as competitive testing); ② Biological verification: Validate the significant differences between different social rank groups predicted by the model in multiple physiological indicators (serum corticosterone, inflammatory factors, gene expression in key brain regions, gut microbiota, etc.); ③ Generalization verification: Apply the model to independent large populations and evaluate its prediction accuracy under different population sizes and feeding environments.

[0030] In some embodiments, for each individual chicken, multiple physiological indicators are determined, including serum corticosterone, inflammatory factors, and gene expression in key brain regions. Based on the serum corticosterone levels of each individual chicken and the predicted social class of each individual chicken, the biological verification of serum corticosterone levels among different social classes of the chicken population was performed. Based on the inflammatory factors of each individual chicken and the predicted social class of each individual chicken, the biological verification of inflammatory factors between different social classes of the chicken population was carried out. Biological verification of key brain region gene expression across different social ranks of the chicken population was conducted based on the gene expression of key brain regions in each individual chicken and the predicted social rank of each individual chicken.

[0031] For example Figure 4As shown, the expression levels of StAR (key brain region gene expression) and CYP11A1 (key brain region gene expression) in the low-grade group (blue) were significantly higher than those in the high-grade group (green). Independent samples t-tests confirmed that the differences were highly significant (P < 0.01).

[0032] Here Figure 4 and Figure 5 In the chart, the P-value represents the probability value in statistical hypothesis testing, used to measure the probability that the currently observed data result is caused solely by random error (coincidence). In scientific research, P < 0.05 is usually used as the criterion for judging whether the result is statistically significant. Specifically, Figure 4 An asterisk (e.g., **) indicates a highly significant real-group difference in the relative expression levels of target genes such as CYP11A1 and StAR between high and low social class groups; Figure 5 The p-value (combined with the coefficient of determination R^2) is used to test whether there is a significant linear correlation between the levels of stress markers such as corticosterone and the expression levels of specific genes (such as P2RX7 or TAP2). Figure 5 In the table, P=0.012 indicates a significant positive correlation, while P=0.48 indicates no significant correlation. In principle, these P values ​​are calculated by first establishing a null hypothesis that there is no difference between the two groups or no correlation between the variables, then selecting an appropriate statistical analysis method (such as an independent samples t-test or linear regression analysis) based on the characteristics of the sample data to calculate the specific statistic, and finally comparing it with the standard probability distribution curve. This helps to objectively determine the reliability of the experimental conclusions.

[0033] like Figure 5 As shown, in the low-grade population, the expression of the pro-inflammatory receptor P2RX7 (inflammatory factor) showed an increasing trend (approximately 1.7-fold) with the increase of serum corticosterone; while in the high-grade population, the expression of TAP2, which is related to antigen presentation, was dominant.

[0034] The biological verification of serum corticosterone is similar to the above-mentioned biological verification methods. It mainly verifies that there is a significant difference between the serum corticosterone of high-grade domestic chickens and low-grade domestic chickens, thereby verifying the accuracy of the predicted social class of this application.

[0035] 1. Early Warning Behavioral Characteristic Indicator System The key quantitative features of the early warning behavior extracted in this application include at least one or more of the following: Response latency: the time interval from the emission of a standardized auditory stimulus to the first recognizable action performed by an individual chicken; Percentage of time spent in resource areas: e.g. Figure 6As shown, this represents the proportion of time an individual chicken spends in the core resource area (such as the foraging area) to the total observation time (the preset time described in this application); Movement intensity and trajectory: The distance a domestic chicken moves within a unit of time (e.g., the preset time described in this application) after being stimulated by a standardized sound; Frequency of risk avoidance behavior: the frequency with which an individual exhibits behaviors such as marginalization and avoidance (e.g., moving away from resource areas or away from the center of the chicken population).

[0036] 2. Standardized test conditions Test environment: A quiet, controlled, standardized environment, free from interference from non-experimental factors; Stimulation method: Uniform sound stimulation parameters (preset sound intensity, preset sound frequency, preset sound duration); Recording method: Multi-angle high-definition video recording to ensure that each individual chicken can be uniquely identified and tracked.

[0037] 3. Two-factor authentication framework This application adopts a dual validation framework of "behavioral prediction + multi-system physiological indicators": Behavioral level: The model output of the ranking prediction results is compared with the traditional gold standard method (validation set, which is the real social ranking of each domestic chicken individual by human labeling); Physiological level: By detecting indicators such as serum corticosterone (HPA axis stress marker), inflammatory factors (immune status), expression of stress-related genes in key brain regions (StAR, CYP11A1, etc.), and gut microbiota structure, we verified that there are significant differences between different prediction grade groups that are consistent with biological expectations.

[0038] In this application, the specific operation process and calculation logic of the social rank prediction model are as follows: After completing the quantification of behavioral characteristics, each individual chicken obtains a comprehensive behavioral characteristic vector V composed of four core indicators. i = [L i norm ,P i , I i , F i Using the four behavioral indicators mentioned above as independent variables (model input) and the social ranking obtained in step 1 using the traditional gold standard method as the dependent variable (target output for model training), a prediction model is established using algorithms such as multiple linear regression or random forest regression. Taking multiple linear regression as an example, the basic form of the model is Rank i = β0 + β1·L i norm + β2·P i + β3·I i + β4·F i, where Rank i Let be the predicted rank value of the i-th individual (the smaller the value, the higher the rank), and each β coefficient is determined by fitting the training data.

[0039] For example, suppose we fit the regression equation Rank to a population of 8 chickens. i = 10 - 2.5×L i norm - 3.2×P i + 1.8×I i + 2.1×F i Existing individual A (L) norm Substituting (=0.2, P=75%, I=8 cm / s, F=0.5 times / minute), we get Rank_A≈22.55, individual B (L norm Substituting the values ​​(e.g., P=1.5, I=25%, I=18 cm / s, F=3.5 times / minute) into the equation, we get Rank_B≈45.20. Since a larger Rank value indicates a lower rank, the predicted rank of individual A is higher than that of individual B. By sorting the Rank values ​​of all individuals in the group from smallest to largest, a complete social rank sequence can be output. This predicted rank is then verified through accuracy and biological validation in step 5 to ensure that the model output rank is highly consistent with the gold standard rank, and that there are significant differences between the high and low rank groups in physiological indicators such as HPA axis stress level and immune function, as expected.

[0040] In some embodiments, after obtaining the Rank value (predicted rank score), social ranks can be divided according to the predicted rank score. For example, the first preset score range can be 0 to 24.9, the second preset score range can be 25 to 49.9, and the third preset score range is 50 or above, so as to determine the social rank of each individual chicken.

[0041] like Figure 7 and Figure 9 As shown, the behavioral feature extraction and quantification process specifically includes the calculation of the following four core indicators: reaction latency ( ( ) is defined as the time difference between the emission of a stimulus and the individual's first identifiable response, i.e. , Let be the moment when the i-th domestic chicken individual first makes a recognizable response (action) after being stimulated by a standardized sound. This refers to the standardized timing of sound stimulus emission.

[0042] The latency period of unresponsive individuals is recorded as the upper limit of the observation window (e.g., 300 seconds), and further observation can be performed. Score standardization was used to eliminate baseline differences; the proportion of time spent in resource areas ( The percentage of time an individual spends within the resource zone of the target area relative to the total observation time (e.g., the preset time) is denoted as _____. ,in This is the cumulative time period for multiple entries and exits from the resource zone. Let be the duration of the i-th domestic chicken individual's stay in the resource area. Total observation time.

[0043] Exercise intensity ( () represents the cumulative distance traveled per unit time. ,in , The centroid coordinates of the individual in each video frame.

[0044] Frequency of risk-avoidance behavior ( This represents the number of times a pre-defined risk-avoidance behavior (such as fleeing resource areas or moving away from the group center) will occur per unit of time. , The number of times risk-avoidance behaviors (such as fleeing resource areas, moving away from the center of the group, etc.) occur; The above features are quantified and integrated into a comprehensive behavioral feature vector for an individual. As the input independent variable of the subsequent prediction model, the social rank prediction model is established with social rank ranking as the dependent variable.

[0045] In summary, this application has the following beneficial effects: Fast and efficient: A single sound stimulus plus a 5-minute video analysis can complete the level assessment, which is more than 95% shorter than the traditional observation time of several weeks.

[0046] Biological validation is reliable: The high and low grade groups predicted by the model showed extremely significant differences in HPA axis stress genes (StAR, CYP11A1, P<0.01), spleen immune indicators (P2RX7 upregulated by 1.7-fold), and behavioral indicators (resource zone dwell time and response latency, P<0.001). Furthermore, adrenal stress and spleen inflammation were extremely significantly positively correlated (P<0.001), demonstrating that the predicted grades have real physiological significance.

[0047] Objective quantification: Features are automatically extracted based on computer vision, eliminating subjective bias from human observation, and the results are reproducible.

[0048] Non-invasive and interference-free: No need to capture, tag, or wear sensors; simply record from a distance using a camera, avoiding additional stress.

[0049] Standardization of conditions: Standardize stimulus parameters and environmental conditions to ensure comparability of results across batches and sites.

[0050] It has great potential for generalization: it can be predicted and validated in large-scale domestic chicken populations, has low hardware costs, and is suitable for large-scale breeding and promotion.

[0051] According to one aspect of the embodiments of this application, a rapid prediction system for the social hierarchy of domestic chickens based on early warning behavior is also proposed. The rapid prediction system 400 for the social hierarchy of domestic chickens based on early warning behavior is as follows: Figure 10 As shown, the system includes: The data acquisition unit 401 is used to conduct a standardized early warning behavior experiment on a group of chickens in the target area during the model training phase, and to collect the reaction characteristics of each individual chicken in the chicken group within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area within the target area, the movement intensity, and the frequency of risk avoidance behavior. Integration unit 402 is used to integrate the reaction latency, the proportion of time spent in the resource area, the movement intensity, and the frequency of risk avoidance behavior of each chicken individual within a preset time period into a comprehensive behavioral feature vector of the chicken individual. Training unit 403 is used to input the comprehensive behavioral feature vector of each of the domestic chickens as training samples into the preset social ranking prediction model, and to train the social ranking prediction model using a pre-built validation set to obtain the trained social ranking prediction model. The acquisition unit 404 is used to acquire the fitting coefficients for the response features output by the trained social rank prediction model; The prediction unit 405 is used in the inference phase to determine the predicted level score of each chicken individual based on the comprehensive behavioral feature vector of the chicken individual and the fitting coefficient output by the trained social level prediction model, and to determine the predicted social level of the chicken individual based on the predicted level score.

[0052] The above is a detailed description of the embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. A method for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior, characterized in that, The method includes: During the model training phase, a standardized early warning behavior experiment is conducted on a group of domestic chickens in the target area. The reaction characteristics of each individual domestic chicken in the group are collected within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area within the target area, the intensity of movement, and the frequency of risk avoidance behavior. For each individual chicken, the chicken's reaction latency, the proportion of time spent in the resource area, the intensity of movement, and the frequency of risk avoidance behavior within the preset time period are integrated into a comprehensive behavioral feature vector for the chicken. The comprehensive behavioral feature vectors of each of the domestic chicken individuals are used as training samples and input into the preset social ranking prediction model. The social ranking prediction model is trained using a pre-constructed validation set to obtain the trained social ranking prediction model. Obtain the fitting coefficients for the response features output by the trained social rank prediction model; During the inference phase, for each individual chicken, the predicted level score of the chicken is determined based on the comprehensive behavioral feature vector of the chicken and the fitting coefficient output by the trained social level prediction model, and the predicted social level of the chicken is determined based on the predicted level score of the chicken.

2. The method for rapid prediction of social hierarchy in domestic chickens based on early warning behavior according to claim 1, characterized in that, The standardized early warning behavior experiment on domestic chickens within the target area includes: A standardized sound stimulus consisting of a preset sound intensity, a preset sound frequency, and a preset sound duration is applied to the chicken population within the target area to provide sound stimulation to each chicken population.

3. The method for rapid prediction of social hierarchy in domestic chickens based on early warning behavior according to claim 2, characterized in that, The reaction latency is the time interval from the emission of the standardized sound stimulus to the individual chicken taking action; the resource area is located at the center of the target area; the dwell time ratio is the proportion of the time the individual chicken spends in the resource area to the preset time; the movement intensity is used to characterize the distance the individual chicken moves within the preset time; the risk avoidance behavior frequency is the number of times the individual chicken moves away from the resource area or the center of the chicken group within the preset time.

4. The method for rapid prediction of social hierarchy in domestic chickens based on early warning behavior according to claim 3, characterized in that, The step of inputting the comprehensive behavioral feature vectors of each of the chicken individuals as training samples into a preset social ranking prediction model, and training the social ranking prediction model using a pre-constructed validation set to obtain the trained social ranking prediction model includes: For each individual chicken, the comprehensive behavioral feature vector of the individual chicken is input into the social rank prediction model to obtain the predicted rank score corresponding to the individual chicken output by the social rank prediction model. The predicted rank score is calculated by the fitting coefficient currently output by the social rank prediction model and the preset fitting formula. The predicted social rank of the individual chicken is determined based on the predicted rank score. In the validation set, the true social class of the chicken individual is determined, and the loss value corresponding to the chicken individual is determined based on the quantitative difference between the predicted social class and the true social class of the chicken individual. The social ranking prediction model is trained based on the loss value corresponding to each individual chicken to iteratively optimize the fitting coefficients and obtain the trained social ranking prediction model.

5. The method for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior according to claim 4, characterized in that, The preset fitting formula is expressed by the following formula: Rank i = β0 + β1·L i norm + β2·P i + β3·I i + β4·F i In the formula, Rank i The predicted grade score for the i-th domestic chicken individual, β0, β1, β2, β3, and β4 constitute the fitting coefficients, L i norm Let P be the latency period of the reaction of the i-th domestic chicken individual. i I represents the proportion of time the i-th domestic chicken individual spends in the resource area. i Let F be the exercise intensity of the i-th domestic chicken individual. i Let be the frequency of the risk avoidance behavior of the i-th domestic chicken individual.

6. The method for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior according to claim 5, characterized in that, During the inference phase, for each individual chicken, a predicted grade score is determined based on the individual chicken's comprehensive behavioral feature vector and the fitting coefficients output by the trained social rank prediction model, including: During the inference phase, for each individual chicken, the response latency, the proportion of time spent in the resource area within the target region, the movement intensity, and the frequency of risk avoidance behavior in the comprehensive behavioral feature vector of the individual chicken, as well as the fitting coefficients output by the trained social rank prediction model, are input into the fitting formula to obtain the predicted rank score of the individual chicken.

7. The method for rapid prediction of the social hierarchy of domestic chickens based on early warning behavior according to claim 6, characterized in that, The step of determining the predicted social class of a chicken individual based on its predicted class score includes: If the predicted social class score of the individual chicken is within the first preset score range, then the predicted social class of the individual chicken is determined to be high. If the predicted social grade of the individual chicken is within the second preset score range, then the predicted social grade of the individual chicken is determined to be intermediate. If the predicted social grade score of the individual chicken is within the third preset score range, then the predicted social grade of the individual chicken is determined to be low. Wherein, the maximum value in the first preset scoring range is less than the minimum value in the second preset scoring range, and the maximum value in the second preset scoring range is less than the minimum value in the third preset scoring range.

8. The method for rapid prediction of social hierarchy in domestic chickens based on early warning behavior according to claim 7, characterized in that, After determining the predicted social class of the chicken individual based on its predicted class score, the method further includes: For each individual chicken, multiple physiological indicators were determined, including serum corticosterone, inflammatory factors, and gene expression in key brain regions. Based on the serum corticosterone levels of each individual chicken and the predicted social class of each individual chicken, the biological verification of serum corticosterone levels among different social classes of the chicken population was performed. Based on the inflammatory factors of each individual chicken and the predicted social class of each individual chicken, the biological verification of inflammatory factors between different social classes of the chicken population was carried out. Biological verification of key brain region gene expression across different social ranks of the chicken population was conducted based on the gene expression of key brain regions in each individual chicken and the predicted social rank of each individual chicken.

9. A rapid prediction system for the social hierarchy of domestic chickens based on early warning behavior, characterized in that, The system includes: The data acquisition unit is used to conduct a standardized early warning behavior experiment on a group of chickens in the target area during the model training phase, and to collect the reaction characteristics of each individual chicken in the chicken group within a preset time period. The reaction characteristics include the reaction latency, the proportion of time spent in the resource area within the target area, the movement intensity, and the frequency of risk avoidance behavior. The integration unit is used to integrate the response latency, the proportion of time spent in the resource area, the movement intensity, and the frequency of risk avoidance behavior of each chicken individual within a preset time period into a comprehensive behavioral feature vector of the chicken individual. The training unit is used to input the comprehensive behavioral feature vector of each of the chicken individuals as training samples into the preset social ranking prediction model, and to train the social ranking prediction model using a pre-built validation set to obtain the trained social ranking prediction model. An acquisition unit is used to acquire the fitting coefficients for the response features output by the trained social rank prediction model. The prediction unit is used in the inference phase to determine the predicted level score of each chicken individual based on the comprehensive behavioral feature vector of the chicken individual and the fitting coefficient output by the trained social level prediction model, and to determine the predicted social level of the chicken individual based on the predicted level score.

Images