Livestock and poultry health monitoring and early warning method and system based on behavior analysis

Abstract

The present application relates to the technical field of livestock health monitoring and early warning, in particular to a livestock health monitoring and early warning method and system based on behavior analysis. The present application clusters disease behavior detection data and non-disease behavior detection data leading to a decrease in livestock yield or quality; establishes a feedback data analysis model to analyze disease behavior detection data, non-disease behavior detection data and positive feedback monitoring parameters, generates first abnormal breeding monitoring parameters and second abnormal breeding monitoring parameters; through real-time monitoring of livestock breeding environment, the similarity with the first abnormal breeding monitoring parameters and the second abnormal breeding monitoring parameters is calculated for corresponding early warning; through this method, historical abnormal breeding parameters can be summarized and analyzed; it is convenient to early warn these abnormal data when monitoring or adjusting the breeding environment monitoring parameters subsequently, reduce the workload of relevant personnel, and improve the accuracy of breeding environment monitoring parameter research.

Description

Technical Field

[0001] This invention relates to the field of livestock and poultry health monitoring and early warning technology, specifically to a livestock and poultry health monitoring and early warning method and system based on behavioral analysis. Background Technology

[0002] Large-scale farming is the main development direction of livestock and poultry farming. The increased stocking density makes it easy for diseases to spread rapidly once an individual in the herd becomes ill, causing large-scale economic losses. In the current farming model, livestock and poultry health monitoring mostly relies on manual inspections, which not only requires a large investment of manpower but also makes it difficult to detect abnormalities in real time. Often, the disease is only detected when obvious symptoms appear, missing the best time for intervention.

[0003] Many existing monitoring technologies focus on abnormalities in activity patterns, feeding, and social behavior. However, these abnormalities, besides being caused by disease, can also be due to comfort issues caused by the environment, and misjudgments can often occur because environmental data is overlooked. In addition, existing monitoring solutions generally lack intelligent automatic early warning capabilities and cannot analyze and interpret the collected behavioral data, making it difficult to meet the needs of precision health management in modern large-scale farming.

[0004] To address this, a method and system for monitoring and early warning of livestock and poultry health based on behavioral analysis is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for monitoring and early warning of livestock and poultry health based on behavior analysis. By summarizing historical data and analyzing and storing abnormal breeding parameters, it can automatically avoid these abnormal data when monitoring or adjusting breeding environment monitoring parameters in subsequent processes, thereby improving the work efficiency of relevant personnel and the accuracy of research on breeding environment monitoring parameters.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] Behavioral analysis-based methods for livestock and poultry health monitoring and early warning include:

[0008] Obtain negative feedback monitoring parameters and positive feedback monitoring parameters of the livestock and poultry breeding environment;

[0009] Clustering and classification are performed based on the negative feedback results of the positive feedback monitoring parameters and the negative feedback monitoring parameters; disease behavior detection data and non-disease behavior detection data are generated, and a feedback data analysis model is established.

[0010] Optionally, the feedback data analysis model includes a feedback data processing unit, a feedback data feature extraction unit, a first abnormal aquaculture monitoring parameter analysis unit, a second abnormal aquaculture monitoring parameter analysis unit, an abnormal data output storage unit, and an abnormal similarity calculation unit;

[0011] The feedback data processing unit filters the disease behavior detection data, non-disease behavior detection data, and positive feedback monitoring parameters according to livestock and poultry breeds and feeding types, and preprocesses the data to generate preprocessed feedback data.

[0012] The feedback data feature extraction unit extracts features from the preprocessed feedback data based on Transformer, GRU, dilated convolution and feature fusion layer to generate behavioral feedback feature data and environmental feedback feature data.

[0013] The first abnormal aquaculture monitoring parameter analysis unit generates the first abnormal aquaculture monitoring parameters based on the abnormal feature data of behavioral feedback feature data and environmental feedback feature data; the second abnormal aquaculture monitoring parameter analysis unit generates the second abnormal aquaculture monitoring parameters by comparing the behavioral feedback feature data or environmental feedback feature data with the non-disease behavior environmental feedback feature data in the knowledge base.

[0014] The abnormal data output storage unit outputs and stores the first and second abnormal breeding monitoring parameters into the corresponding database. This method can effectively distinguish abnormal data caused by disease behavior from normal breeding activities by combining feedback parameters with behavioral feature extraction, reducing the interference of non-disease factors such as environmental fluctuations and feeding operations on monitoring results, improving the accuracy of livestock and poultry health abnormality identification, and quickly locating potential health risks through abnormality similarity matching, achieving early warning and early intervention for livestock and poultry diseases, and reducing breeding losses caused by disease spread.

[0015] The disease behavior detection data and the non-disease behavior detection data are analyzed based on the feedback data analysis model to generate and store the first abnormal breeding monitoring parameters;

[0016] Optionally, the first abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model; in this embodiment, the first abnormal aquaculture monitoring parameter is an abnormal aquaculture monitoring parameter that is prone to causing disease.

[0017] The behavioral feedback feature data or environmental feedback feature data is divided equally according to the sliding time window to obtain multiple feedback feature data, wherein the feedback feature data includes at least environmental trend feature data and morphological feature data.

[0018] The t-th feedback feature data and the (t-1)-th feedback feature data are used to perform a first feedback similarity calculation. This first feedback similarity calculation includes trend feature similarity calculation and morphological feature similarity calculation. If the result of the first feedback similarity calculation is less than the first feedback similarity threshold, then the t-th feedback feature data is the first abnormal livestock monitoring parameter. The trend feature similarity calculation includes at least the standard deviation method, empirical statistical method, and dynamic path calculation method; the morphological feature similarity calculation includes at least the feature point matching method and image recognition analysis method. This method can capture early abnormal changes in livestock behavior. Compared with traditional monitoring methods, it can identify potential health risks in advance, reduce the probability of large-scale disease outbreaks, reduce the labor costs of manual inspections, and improve the intelligence level and early warning accuracy of livestock monitoring. By distinguishing between trend features and morphological features in hierarchical similarity calculation, it takes into account both the development trend and immediate manifestation of behavioral changes, effectively reducing the probability of misjudgment and missed judgment, and ensuring the reliability of monitoring results.

[0019] The expressions for calculating the trend feature similarity and the morphological feature similarity are as follows:

[0020] ;

[0021] ;

[0022] in, The result of the similarity calculation for the trend features. The standard deviation method, The empirical statistical method is described above. This refers to the dynamic path calculation method. For the t-th environmental trend feature data, For the (t-1)th environmental trend feature data, , and The weights calculated for the similarity of the trend features. ; The result is the morphological feature similarity calculation, and C represents the transformation of heatmap data. For ORB similarity algorithm, The image recognition and analysis method, For the t-th morphological feature data, For the (t-1)th morphological feature data, and The weights calculated for the morphological feature similarity; This method adapts to the monitoring needs of different breeding scenarios and livestock breeds by pre-setting weights, and flexibly adjusts the proportion of trend features and morphological features in similarity calculation, further improving the adaptability and accuracy of anomaly identification. This allows the monitoring model to match the actual application needs of farms of different sizes, providing reliable data support for subsequent accurate early warning.

[0023] Based on the feedback data analysis model, the disease behavior detection data, the non-disease behavior detection data, and the knowledge base data are compared to generate and store the second abnormal aquaculture monitoring parameters;

[0024] Optionally, the second abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model;

[0025] The behavioral feedback feature data, environmental feedback feature data, and non-disease behavior detection data are equally divided according to a sliding time window to obtain multiple real-time feedback detection data and multiple non-disease behavior detection data. The real-time feedback detection data includes at least environmental trend feature data and morphological feature data; the non-disease behavior detection data includes at least non-disease trend data and non-disease morphological data. A second feedback similarity calculation is performed between the k-th real-time feedback detection data and the k-th non-disease behavior detection data. If the result of the second feedback similarity calculation is greater than the second feedback similarity threshold, then the k-th real-time feedback detection data is the second abnormal aquaculture monitoring parameter.

[0026] The expression for calculating the second feedback similarity is as follows:

[0027] ;

[0028] in, The result of the second feedback similarity calculation. The empirical statistical method is described above. This refers to the dynamic path calculation method. For the kth environmental trend feature data, For the k-th non-disease trend data, For the kth morphological feature data, Let C be the k-th non-disease morphology data, and let C be the heatmap data transformation. For ORB similarity algorithm, The image recognition and analysis method, , , and The weights calculated for the similarity of the trend features. This method can capture behavioral anomalies and environmental data anomalies in livestock and poultry behavior by comparing them with a knowledge base, avoiding the judgment errors caused by a single similarity calculation method and improving the identification accuracy of the second abnormal breeding monitoring parameter. At the same time, by using a sliding window equal division matching method, the positioning range of abnormal parameters is narrowed, which can more quickly pinpoint the time period when behavioral anomalies occur, providing more accurate abnormal data support for subsequent livestock and poultry health early warning, reducing the probability of missed or false detection of health anomalies, and ensuring the reliability of monitoring and early warning.

[0029] The livestock and poultry breeding environment is monitored in real time, and the livestock and poultry breeding environment is compared with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model.

[0030] If the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold, an aquaculture monitoring early warning will be issued for adjustment.

[0031] If the first or second abnormal aquaculture monitoring parameter adjusted by relevant personnel is not less than the aquaculture environment similarity threshold, a secondary aquaculture monitoring warning will be issued.

[0032] This application also provides a livestock and poultry health monitoring and early warning system based on behavior analysis, including:

[0033] The negative feedback data acquisition module is used to acquire negative feedback monitoring parameters of the livestock and poultry breeding environment;

[0034] The positive feedback data acquisition module is used to acquire positive feedback monitoring parameters of the livestock and poultry breeding environment;

[0035] The negative feedback data classification module is used to perform clustering and classification based on the positive feedback monitoring parameters and the negative feedback results of the negative feedback monitoring parameters; and to generate disease behavior detection data and non-disease behavior detection data.

[0036] The negative feedback data analysis module is used to establish a feedback data analysis model, analyze the disease behavior detection data and the non-disease behavior detection data based on the feedback data analysis model, generate and store the first abnormal breeding monitoring parameters; and compare the disease behavior detection data, the non-disease behavior data and the knowledge base data based on the feedback data analysis model to generate and store the second abnormal breeding monitoring parameters.

[0037] The real-time monitoring module for the breeding environment is used to monitor the livestock and poultry breeding environment in real time, and compare the livestock and poultry breeding environment with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model.

[0038] A primary aquaculture monitoring and early warning module is used to issue a primary aquaculture monitoring early warning for adjustment if the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold.

[0039] The secondary aquaculture monitoring and early warning module is used to issue a secondary aquaculture monitoring early warning if the first or second abnormal aquaculture monitoring parameter adjusted by relevant personnel is not less than the aquaculture environment similarity threshold.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1. By simultaneously combining primary and secondary abnormal breeding monitoring parameters for dual anomaly detection, this method can capture abnormal changes in livestock and poultry behavior over time and identify potential health problems through horizontal comparison of feedback parameters. This significantly improves the accuracy of livestock and poultry disease monitoring and reduces the probability of misjudgments and missed diagnoses caused by single-dimensional monitoring. Furthermore, relying on sliding time windows and feature similarity calculation methods, it enables early identification of anomalies, issuing warnings at the initial stage of disease outbreaks, allowing sufficient intervention time for farmers and reducing economic losses caused by large-scale livestock and poultry disease. Based on the multi-feature extraction and fusion of Transformer, GRU, and dilated convolution, it can also adapt to the behavioral data features of livestock and poultry under different breeds and feeding models, offering stronger adaptability and wide applicability to intelligent health monitoring scenarios in large-scale farms. Attached Figure Description

[0042] Figure 1 This is a flowchart of the method of the present invention.

[0043] Figure 2 This is a flowchart of the feedback data analysis model of the present invention.

[0044] Figure 3 This is a network structure diagram of the feedback data feature extraction unit of the feedback data analysis model of the present invention.

[0045] Figure 4 This is a system flowchart of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] This invention provides a method and system for monitoring and early warning of livestock and poultry health based on behavior analysis. Taking the health monitoring of laying hens as an example, the technical solution is as follows:

[0048] Example 1:

[0049] A method for monitoring and early warning of livestock and poultry health based on behavioral analysis, referring to Figure 1 As shown, it includes.

[0050] Acquire negative feedback monitoring parameters and positive feedback monitoring parameters for the livestock and poultry farming environment; positive feedback monitoring parameters (e.g., but not limited to temperature data, humidity data, light duration data, carbon dioxide concentration data, ammonia concentration data, etc.); the environmental data of the positive feedback monitoring parameters are time-series data;

[0051] Clustering and classification are performed based on the positive feedback monitoring parameters and the negative feedback results of the negative feedback monitoring parameters; disease behavior detection data and non-disease behavior detection data are generated, and a feedback data analysis model is established; the disease behavior detection data is negative feedback data caused by chicken plague in laying hens, and the negative feedback data includes environmental data (e.g., but not limited to temperature data, humidity data, light duration data, carbon dioxide concentration data, ammonia concentration data, etc. during chicken plague) and behavioral data (e.g., images of laying hen activity, feeding images, resting images, etc.).

[0052] The non-disease behavior detection data includes environmental data (e.g., but not limited to temperature data, humidity data, light duration data, carbon dioxide concentration data, and ammonia concentration data during non-chicken plague-related abnormalities in egg production and quality) and behavioral data (e.g., images of laying hens' activities, feeding images, and resting images) and behavioral data (e.g., images of laying hens' activities, feeding images, and resting images) and time-series data for negative feedback data on non-chicken plague-related abnormalities in egg production and quality.

[0053] The negative feedback data caused by chicken plague in laying hens and the behavioral, environmental, and image data of laying hens without chicken plague are associated with each other through labels;

[0054] Specifically, using existing environmental data of negative feedback monitoring parameters, relevant technicians mark the negative feedback monitoring parameters when fowl plague occurs and the negative feedback monitoring parameters when fowl plague does not occur; using temperature data as the horizontal axis and other negative feedback parameters as the vertical axis, a feedback breeding parameter coordinate system is generated. The other negative feedback parameters can only be of one type, for example, temperature as the horizontal axis and humidity as the vertical axis; or temperature as the horizontal axis and light duration data as the vertical axis.

[0055] The parameters during the occurrence of fowl plague were clustered using the density clustering algorithm (DBSCAN algorithm) to determine the clustering range of negative feedback parameters for various fowl plagues; the positive feedback monitoring parameters were clustered to generate the clustering range of various positive feedback parameters; and the negative feedback parameters when fowl plague did not occur were clustered to determine the clustering range of negative feedback parameters for various fowl plagues.

[0056] By comparing the clustering range of feedback parameters during fowl plague with the clustering range of positive feedback parameters, the different clustering ranges are used to determine the clustering ranges of various key negative feedback parameters when fowl plague occurs. Similarly, by comparing the clustering range of negative feedback parameters when fowl plague does not occur with the clustering ranges of positive feedback parameters and the clustering ranges of key negative feedback parameters during fowl plague, the different clustering ranges are used to determine the clustering ranges of various key negative feedback parameters when fowl plague does not occur. By obtaining image data with corresponding associated labels based on the clustering results of environmental data, image data for both fowl plague and non-fowl plague scenarios can be obtained.

[0057] Optionally, refer to Figure 2 As shown, the feedback data analysis model includes a feedback data processing unit, a feedback data feature extraction unit, a first abnormal aquaculture monitoring parameter analysis unit, a second abnormal aquaculture monitoring parameter analysis unit, an abnormal data output storage unit, and an abnormal similarity calculation unit.

[0058] The feedback data processing unit filters the disease behavior detection data, non-disease behavior detection data, and positive feedback monitoring parameters according to livestock breeds and feeding types (e.g., but not limited to floor rearing, cage rearing, etc.), and preprocesses the data to generate preprocessed feedback data; the preprocessed feedback data includes corresponding behavioral image data and environmental data.

[0059] Reference Figure 3 As shown, the feedback data feature extraction unit extracts features from the preprocessed feedback data based on Transformer, GRU, dilated convolution, and feature fusion layer. Transformer corresponds to... Figure 3 The converter, GRU corresponds to Figure 3 The gated recurrent unit has a ReLU feature fusion layer; it generates behavioral feedback feature data and environmental feedback feature data.

[0060] To ensure the accuracy of feature extraction from behavioral and environmental feedback data—two types of abnormal feedback data—the feedback data feature extraction unit generates an anomaly classification model by connecting a fully connected layer and a Softmax layer after the Transformer, GRU, dilated convolution, and feature fusion layers. In this embodiment, based on a portion of untrained aquaculture parameter data obtained by relevant technical personnel, anomaly classification and identification are performed using the anomaly classification model generated by the feedback data feature extraction unit. The data includes normal and abnormal data, totaling four sets. The first two sets are for environmental data identification, and the last two sets are for image data identification. The identification results are shown in Table 1.

[0061] Table 1. Accuracy of identifying abnormal aquaculture parameter data

[0062] Data Number Accuracy of abnormal aquaculture parameter identification EN01 85.22% EN02 87.04% EN03 88.57% EN04 86.89%

[0063] As can be seen from Table 1, the overall accuracy rate of identification is above 85%, and the identification effect of abnormal breeding parameters is good.

[0064] The first abnormal breeding monitoring parameter analysis unit generates the first abnormal breeding monitoring parameter based on the abnormal feature data of behavioral feedback feature data and environmental feedback feature data; the second abnormal breeding monitoring parameter analysis unit generates the second abnormal breeding monitoring parameter by comparing the behavioral feedback feature data or environmental feedback feature data with non-disease behavioral environmental feedback feature data in the knowledge base, which is set based on expert experience; in this embodiment, both the first and second abnormal breeding monitoring parameters include environmental data and laying hen behavioral data caused by environmental abnormalities due to fowl plague, such as abnormal chicken behavior caused only by excessively high temperature or excessively high humidity;

[0065] The abnormal data output storage unit outputs and stores the first and second abnormal breeding monitoring parameters into the corresponding database. This method can effectively distinguish abnormal data caused by disease behavior from normal breeding activities by combining feedback parameters with behavioral feature extraction, reducing the interference of non-disease factors such as environmental fluctuations and feeding operations on monitoring results, improving the accuracy of livestock and poultry health abnormality identification, and quickly locating potential health risks through abnormality similarity matching, achieving early warning and early intervention for livestock and poultry diseases, and reducing breeding losses caused by disease spread.

[0066] The disease behavior detection data and the non-disease behavior detection data are analyzed based on the feedback data analysis model to generate and store the first abnormal breeding monitoring parameters;

[0067] Optionally, the first abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model;

[0068] The behavioral feedback feature data or environmental feedback feature data is divided equally according to the sliding time window to obtain multiple feedback feature data, wherein the feedback feature data includes at least environmental trend feature data and morphological feature data.

[0069] The t-th feedback feature data and the (t-1)-th feedback feature data are compared using a first feedback similarity calculation. This first feedback similarity calculation includes trend feature similarity calculation and morphological feature similarity calculation. If the result of the first feedback similarity calculation is less than a first feedback similarity threshold, then the t-th feedback feature data is the first abnormal aquaculture monitoring parameter. The first feedback similarity threshold is set based on expert experience and defaults to 90%.

[0070] The trend feature similarity calculation includes at least the standard deviation method, empirical statistical method, and dynamic path calculation method; the morphological feature similarity calculation includes at least the feature point matching method and image recognition analysis method. This method can capture early abnormal changes in livestock and poultry behavior. Compared with traditional monitoring methods, it can identify potential health risks in advance, reduce the probability of large-scale disease outbreaks, reduce the labor costs of manual inspections, and improve the intelligence level and early warning accuracy of livestock monitoring. By distinguishing between trend features and morphological features in hierarchical similarity calculation, it takes into account both the development trend and immediate manifestation of behavioral changes, effectively reducing the probability of misjudgment and missed judgment, and ensuring the reliability of monitoring results.

[0071] The expressions for calculating the trend feature similarity and the morphological feature similarity are as follows:

[0072] ;

[0073] ;

[0074] in, The result of the similarity calculation for the trend features. The standard deviation method, The empirical statistical method is described above. This refers to the dynamic path calculation method. For the t-th environmental trend feature data, For the (t-1)th environmental trend feature data, , and The weights calculated for the similarity of the trend features. ; The result is the morphological feature similarity calculation, and C represents the transformation of heatmap data. For ORB similarity algorithm, This refers to the image recognition and analysis method (such as the pre-trained ViT algorithm). For the t-th morphological feature data, For the (t-1)th morphological feature data, and The weights calculated for the morphological feature similarity; This method adapts to the monitoring needs of different breeding scenarios and livestock breeds by pre-setting weights, and flexibly adjusts the proportion of trend features and morphological features in similarity calculation, further improving the adaptability and accuracy of anomaly identification. This allows the monitoring model to match the actual application needs of farms of different sizes, providing reliable data support for subsequent accurate early warning.

[0075] Based on the feedback data analysis model, the disease behavior detection data, the non-disease behavior detection data, and the knowledge base data are compared to generate and store the second abnormal aquaculture monitoring parameters;

[0076] Optionally, the second abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model;

[0077] Based on a sliding time window, behavioral feedback feature data, environmental feedback feature data, and non-disease behavior detection data are equally divided to obtain multiple real-time feedback detection data and multiple non-disease behavior detection data. The real-time feedback detection data includes at least environmental trend feature data and morphological feature data; the non-disease behavior detection data includes at least non-disease trend data and non-disease morphological data. A second feedback similarity calculation is performed between the k-th real-time feedback detection data and the k-th non-disease behavior detection data. The second feedback similarity calculation is a trend feature similarity calculation. If the result of the second feedback similarity calculation is greater than the second feedback similarity threshold, then the k-th real-time feedback detection data is the second abnormal aquaculture monitoring parameter. The second feedback similarity threshold is set based on expert experience and is 90% by default.

[0078] The behavioral feedback feature data, environmental feedback feature data, and non-disease behavior detection data are equally divided according to a sliding time window to obtain multiple real-time feedback detection data and multiple non-disease behavior detection data. The real-time feedback detection data includes at least environmental trend feature data and morphological feature data; the non-disease behavior detection data includes at least non-disease trend data and non-disease morphological data. A second feedback similarity calculation is performed between the k-th real-time feedback detection data and the k-th non-disease behavior detection data. If the result of the second feedback similarity calculation is greater than the second feedback similarity threshold, then the k-th real-time feedback detection data is the second abnormal aquaculture monitoring parameter.

[0079] The expression for calculating the second feedback similarity is as follows:

[0080] ;

[0081] in, The result of the second feedback similarity calculation. The empirical statistical method is described above. This refers to the dynamic path calculation method. For the kth environmental trend feature data, For the k-th non-disease trend data, For the kth morphological feature data, Let C be the k-th non-disease morphology data, and let C be the heatmap data transformation. For ORB similarity algorithm, This refers to the image recognition and analysis method (such as the pre-trained ViT algorithm). , , and The weights calculated for the similarity of the trend features. This method can capture behavioral anomalies and environmental data anomalies in livestock and poultry behavior by comparing them with a knowledge base, avoiding the judgment errors caused by a single similarity calculation method and improving the identification accuracy of the second abnormal breeding monitoring parameter. At the same time, by using a sliding window equal division matching method, the positioning range of abnormal parameters is narrowed, which can more quickly pinpoint the time period when behavioral anomalies occur, providing more accurate abnormal data support for subsequent livestock and poultry health early warning, reducing the probability of missed or false detection of health anomalies, and ensuring the reliability of monitoring and early warning.

[0082] The abnormal data output storage unit outputs the first abnormal aquaculture monitoring parameter and the second abnormal aquaculture monitoring parameter and stores them into the corresponding database.

[0083] The livestock and poultry breeding environment is monitored in real time, and the livestock and poultry breeding environment is compared with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model.

[0084] If the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold, an aquaculture monitoring early warning will be issued for adjustment.

[0085] If the adjusted first or second abnormal breeding monitoring parameter is not less than the breeding environment similarity threshold, a secondary breeding monitoring warning is issued. If the first breeding monitoring warning identifies fowl plague, an initial adjustment is made based on environmental issues to prevent false alarms. If the warning identifies environmental issues, the initial adjustment is performed directly. The secondary breeding monitoring warning is issued if the warning is still identified as fowl plague, confirming the fowl plague warning. If the warning still identifies environmental issues, it indicates a problem with the adjustment equipment requiring repair. The breeding environment similarity threshold is 90% by default, but can be set by experts based on actual conditions.

[0086] Example 2:

[0087] This application also provides a livestock and poultry health monitoring and early warning system based on behavior analysis, referring to... Figure 4 As shown, it includes:

[0088] The negative feedback data acquisition module is used to acquire negative feedback monitoring parameters of the livestock and poultry breeding environment;

[0089] The positive feedback data acquisition module is used to acquire positive feedback monitoring parameters of the livestock and poultry breeding environment;

[0090] The negative feedback data classification module is used to perform clustering and classification based on the positive feedback monitoring parameters and the negative feedback results of the negative feedback monitoring parameters; and to generate disease behavior detection data and non-disease behavior detection data.

[0091] The negative feedback data analysis module is used to establish a feedback data analysis model, analyze the disease behavior detection data and the non-disease behavior detection data based on the feedback data analysis model, generate and store the first abnormal breeding monitoring parameters; and compare the disease behavior detection data, the non-disease behavior data and the knowledge base data based on the feedback data analysis model to generate and store the second abnormal breeding monitoring parameters.

[0092] The real-time monitoring module for the breeding environment is used to monitor the livestock and poultry breeding environment in real time, and compare the livestock and poultry breeding environment with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model.

[0093] A primary aquaculture monitoring and early warning module is used to issue a primary aquaculture monitoring early warning for adjustment if the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold.

[0094] The secondary aquaculture monitoring and early warning module is used to issue a secondary aquaculture monitoring early warning if the first or second abnormal aquaculture monitoring parameter adjusted by relevant personnel is not less than the aquaculture environment similarity threshold.

[0095] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and early warning of livestock and poultry health based on behavioral analysis, characterized in that, include: Obtain negative feedback monitoring parameters and positive feedback monitoring parameters of the livestock and poultry breeding environment; Clustering and classification are performed based on the negative feedback results of the positive feedback monitoring parameters and the negative feedback monitoring parameters; Generate disease behavior detection data and non-disease behavior detection data, and establish a feedback data analysis model; The disease behavior detection data and the non-disease behavior detection data are analyzed based on the feedback data analysis model to generate and store the first abnormal breeding monitoring parameters; Based on the feedback data analysis model, the disease behavior detection data, the non-disease behavior detection data, and the knowledge base data are compared to generate and store the second abnormal aquaculture monitoring parameters; The livestock and poultry breeding environment is monitored in real time, and the livestock and poultry breeding environment is compared with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model. If the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold, an aquaculture monitoring early warning will be issued for adjustment. If the first or second abnormal aquaculture monitoring parameter adjusted by relevant personnel is not less than the aquaculture environment similarity threshold, a secondary aquaculture monitoring warning will be issued.

2. The livestock and poultry health monitoring and early warning method based on behavior analysis according to claim 1, characterized in that, Clustering and classification are performed based on the negative feedback results of the positive feedback monitoring parameters and the negative feedback monitoring parameters; Generating disease behavior detection data and non-disease behavior detection data, and establishing a feedback data analysis model includes: The feedback data analysis model includes a feedback data processing unit, a feedback data feature extraction unit, a first abnormal aquaculture monitoring parameter analysis unit, a second abnormal aquaculture monitoring parameter analysis unit, an abnormal data output storage unit, and an abnormal similarity calculation unit. The feedback data processing unit filters the disease behavior detection data, non-disease behavior detection data, and positive feedback monitoring parameters according to livestock and poultry breeds and feeding types, and preprocesses the data to generate preprocessed feedback data. The feedback data feature extraction unit extracts features from the preprocessed feedback data based on Transformer, GRU, dilated convolution and feature fusion layer to generate behavioral feedback feature data and environmental feedback feature data. The first abnormal aquaculture monitoring parameter analysis unit generates the first abnormal aquaculture monitoring parameters based on the abnormal feature data of behavioral feedback feature data and environmental feedback feature data; the second abnormal aquaculture monitoring parameter analysis unit generates the second abnormal aquaculture monitoring parameters by comparing the behavioral feedback feature data or environmental feedback feature data with the non-disease behavior environmental feedback feature data in the knowledge base. The abnormal data output storage unit outputs the first abnormal aquaculture monitoring parameter and the second abnormal aquaculture monitoring parameter and stores them in the corresponding database.

3. The livestock and poultry health monitoring and early warning method based on behavior analysis according to claim 2, characterized in that, The disease behavior detection data and the non-disease behavior detection data are analyzed based on the feedback data analysis model to generate and store the first abnormal breeding monitoring parameters; The first abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model, including: The behavioral feedback feature data or environmental feedback feature data is divided equally according to the sliding time window to obtain multiple feedback feature data, wherein the feedback feature data includes at least environmental trend feature data and morphological feature data. The t-th feedback feature data and the (t-1)-th feedback feature data are used to perform a first feedback similarity calculation; the first feedback similarity calculation includes trend feature similarity calculation and morphological feature similarity calculation; If the first feedback similarity calculation result is less than the first feedback similarity threshold, then the t-th feedback feature data is the first abnormal aquaculture monitoring parameter. The trend feature similarity calculation includes at least the standard deviation method, empirical statistical method, and dynamic path calculation method; the morphological feature similarity calculation includes at least the feature point matching method and image recognition analysis method.

4. The livestock and poultry health monitoring and early warning method based on behavior analysis according to claim 3, characterized in that, The expressions for calculating the trend feature similarity and the expressions for calculating the morphological feature similarity include: ； ； in, The result of the similarity calculation for the trend features. The standard deviation method, The empirical statistical method is described above. This refers to the dynamic path calculation method. For the t-th environmental trend feature data, For the (t-1)th environmental trend feature data, , and The weights calculated for the similarity of the trend features. ; The result is the morphological feature similarity calculation, and C represents the transformation of heatmap data. For ORB similarity algorithm, The image recognition and analysis method is described above. For the t-th morphological feature data, For the (t-1)th morphological feature data, and The weights calculated for the morphological feature similarity; .

5. The livestock and poultry health monitoring and early warning method based on behavior analysis according to claim 2, characterized in that, Based on the feedback data analysis model, the disease behavior detection data, the non-disease behavior detection data, and the knowledge base data are compared to generate and store the second abnormal aquaculture monitoring parameters; The second abnormal aquaculture monitoring parameter is calculated and obtained based on the abnormal similarity calculation unit of the feedback data analysis model, including: The behavioral feedback feature data, environmental feedback feature data, and non-disease behavior detection data are equally divided according to a sliding time window to obtain multiple real-time feedback detection data and multiple non-disease behavior detection data. The real-time feedback detection data includes at least environmental trend feature data and morphological feature data; the non-disease behavior detection data includes at least non-disease trend data and non-disease morphological data. A second feedback similarity calculation is performed between the k-th real-time feedback detection data and the k-th non-disease behavior detection data. If the result of the second feedback similarity calculation is greater than the second feedback similarity threshold, then the k-th real-time feedback detection data is the second abnormal aquaculture monitoring parameter. The expression for calculating the second feedback similarity is as follows: ； in, The result of the second feedback similarity calculation. The empirical statistical method is described above. This refers to the dynamic path calculation method. For the kth environmental trend feature data, For the k-th non-disease trend data, For the kth morphological feature data, Let C be the k-th non-disease morphology data, and let C be the heatmap data transformation. For ORB similarity algorithm, The image recognition and analysis method is described above. , , and The weights calculated for the similarity of the trend features. .

6. A livestock and poultry health monitoring and early warning system based on behavioral analysis, characterized in that, include: The negative feedback data acquisition module is used to acquire negative feedback monitoring parameters of the livestock and poultry breeding environment; The positive feedback data acquisition module is used to acquire positive feedback monitoring parameters of the livestock and poultry breeding environment; The negative feedback data classification module is used to perform clustering and classification based on the positive feedback monitoring parameters and the negative feedback results of the negative feedback monitoring parameters; and to generate disease behavior detection data and non-disease behavior detection data. The negative feedback data analysis module is used to establish a feedback data analysis model, analyze the disease behavior detection data and the non-disease behavior detection data based on the feedback data analysis model, generate and store the first abnormal breeding monitoring parameters. Based on the feedback data analysis model, the disease behavior detection data, the non-disease behavior detection data, and the knowledge base data are compared to generate and store the second abnormal aquaculture monitoring parameters; The real-time monitoring module for the breeding environment is used to monitor the livestock and poultry breeding environment in real time, and compare the livestock and poultry breeding environment with the first abnormal breeding monitoring parameter and the second abnormal breeding monitoring parameter through the feedback data analysis model. A primary aquaculture monitoring and early warning module is used to issue a primary aquaculture monitoring early warning for adjustment if the first abnormal aquaculture monitoring parameter or the second abnormal aquaculture monitoring parameter is not less than the aquaculture environment similarity threshold. The secondary aquaculture monitoring and early warning module is used to issue a secondary aquaculture monitoring early warning if the first or second abnormal aquaculture monitoring parameter adjusted by relevant personnel is not less than the aquaculture environment similarity threshold.

Images