Cow health condition intelligent analysis method and system based on feature recognition

Abstract

The invention discloses an intelligent cattle health condition analysis method and system based on feature recognition, and relates to the technical field of intelligent analysis, and the intelligent analysis method comprises the following steps: obtaining cattle health data based on a camera and a sensor which are deployed in advance, and carrying out the preprocessing of the cattle health data; feature recognition is carried out on the preprocessed cattle health data by using an SIFT algorithm, cattle health feature elements are confirmed, and a cattle health assessment feature set is constructed; based on the cattle health assessment feature set and the cattle health data, utilizing an isolated forest algorithm to construct and train a cattle health analysis model, and determining an abnormal score; and the trained cattle health analysis model is utilized to analyze the cattle health data in combination with the abnormal score, and the cattle health state is analyzed and evaluated. The intelligent level of cattle herd health management is greatly improved, and technical support and economic guarantee are provided for development of modern animal husbandry.

Description

Technical Field

[0001] The present invention relates to the field of intelligent analysis technology. Specifically, it relates to an intelligent analysis method and system for the health status of cows based on feature recognition. Background Art

[0002] With the rapid development of modern animal husbandry and the popularization of data technology, data evaluation has been gradually introduced in the process of cow breeding to achieve precise monitoring and management of the health status of cows. However, the traditional methods for cow health assessment have significant limitations and are difficult to meet the requirements of modern pastures for efficient, precise, and real-time management.

[0003] Traditional methods mainly rely on manual observation and the collection of simple physiological indicators (such as body temperature and heart rate). It is difficult to comprehensively reflect the overall health status of cows only relying on simple physiological indicators. Especially in the sub-healthy stage, early abnormal signals are often missed, resulting in the delayed discovery or even deterioration of health problems. This makes traditional methods appear inefficient and unreliable when facing the requirements of modern livestock management.

[0004] At the same time, modern cow health assessment often involves multi-modal data, including image data (such as body posture and gesture), behavior data (such as feeding and activities), physiological data (such as body temperature and heart rate), and environmental data (such as temperature, humidity, and ammonia concentration). However, these data sources are diverse and have different formats. For example, images are high-dimensional visual data, behavior data are time series, while physiological and environmental data are scalar data. The characteristic differences and uneven distributions among these multi-modal data make data fusion and processing extremely complex. In addition, the collection frequencies and precisions of different data may be inconsistent, further increasing the difficulty of integration. Existing technologies are difficult to construct a unified feature representation method and cannot fully explore the correlation between multi-modal data, thus limiting the accuracy and reliability of health assessment. Health detection mainly relies on regular manual inspections, lacking real-time nature, resulting in health abnormalities usually developing into more serious problems or even spreading when discovered, missing the best intervention opportunity, thereby increasing economic losses and health risks.

[0005] Regarding the problems in the related technologies, no effective solutions have been proposed yet. Summary of the Invention

[0006] Regarding the problems in the related technologies, the present invention proposes an intelligent analysis method and system for the health status of cows based on feature recognition to overcome the above-mentioned technical problems existing in the existing related technologies.

[0007] To this end, the specific technical solutions adopted by the present invention are as follows:

[0008] According to one aspect of the present invention, there is provided an intelligent analysis method for the health status of cows based on feature recognition. The intelligent analysis method includes the following steps:

[0009] S1. Obtain cattle health data based on pre-deployed cameras and sensors, and preprocess the cattle health data;

[0010] S2. Use the SIFT algorithm to perform feature recognition on the preprocessed cattle health data, confirm the cattle health feature elements, and construct a cattle health assessment feature set;

[0011] S3. Based on the cattle health assessment feature set and the cattle health data, use the Isolation Forest algorithm to construct and train a cattle health analysis model to determine the anomaly score;

[0012] S4. Use the trained cattle health analysis model to analyze the cattle health data in combination with the anomaly score, and analyze and evaluate the cattle health status.

[0013] Optionally, the obtaining of cattle health data based on pre-deployed cameras and sensors and the preprocessing of the cattle health data include the following steps:

[0014] S11. Collect cattle health data of cattle through pre-deployed cameras and sensors, where the cattle health data includes the daily behavior images and videos of cattle and the physiological data of cattle;

[0015] S12. Perform data cleaning and normalization processing on the collected cattle health data, and convert the processed data into a unified standard format.

[0016] Optionally, the using of the SIFT algorithm to perform feature recognition on the preprocessed cattle health data, confirming the cattle health feature elements, and constructing a cattle health assessment feature set includes the following steps:

[0017] S21. Use the SIFT algorithm to extract the key feature points in the preprocessed cattle health data;

[0018] S22. Based on the cattle health data, combine the extracted key feature points to construct a health assessment feature set.

[0019] Optionally, the using of the SIFT algorithm to extract the key feature points in the preprocessed cattle health data includes the following steps:

[0020] S211. Use the SIFT algorithm to detect the key feature points of the images in the preprocessed cattle health data, and locate the distribution of the key feature points through the construction of a Gaussian pyramid;

[0021] S212. Based on the distribution of the key feature points, generate the feature vectors corresponding to each key feature point, and calculate the gradient distribution around the key feature points.

[0022] Optionally, based on the bovine health assessment feature set and bovine health data, using the Isolation Forest algorithm, constructing and training a bovine health analysis model, and determining the anomaly score includes the following steps:

[0023] S31. Using the Isolation Forest algorithm, combining with the bovine health assessment feature set, constructing a bovine health analysis model, and determining the anomaly score;

[0024] S32. Training and optimizing the bovine health analysis model, and evaluating the performance of the bovine health analysis model.

[0025] Optionally, the using the Isolation Forest algorithm, combining with the bovine health assessment feature set, constructing a bovine health analysis model, and determining the anomaly score includes the following steps:

[0026] S311. Based on the Isolation Forest algorithm, formulating an anomaly score for the bovine health assessment feature set in combination with bovine health data;

[0027] S312. Separating the outliers in the bovine health assessment feature set by random partitioning. If the data in the bovine health assessment feature set is lower than the anomaly score, it is an outlier;

[0028] S313. If the data in the bovine health assessment feature set reaches or exceeds the anomaly score, it is a normal value.

[0029] Optionally, the expression of the anomaly score is:

[0030] s(x,n) = 2 - c(n)E(h(x))

[0031] In the formula, s(x,n) represents the anomaly score, E(h(x)) represents the average separation depth of the data point x in the bovine health assessment feature set on all random trees, c(n) represents the regularization factor of the average separation depth of the random trees, and n represents the number of samples in the bovine health assessment feature set.

[0032] Optionally, using the trained bovine health analysis model, combining with the anomaly score to analyze the bovine health data, and analyzing and evaluating the bovine health status includes the following steps:

[0033] S41. Based on the bovine health analysis model, calculating the anomaly score for the bovine in real time, setting the anomaly threshold according to the anomaly score, and dividing the bovine health status;

[0034] S42. Issuing a warning for the bovine health status, recording the bovines with health anomalies, and reporting them.

[0035] Optionally, the based on the bovine health analysis model, calculating the anomaly score for the bovine in real time, setting the anomaly threshold according to the anomaly score, and dividing the bovine health status includes the following steps:

[0036] S411. Calculate the abnormal score of the cattle in real time based on the cattle health analysis model, and divide the health status of the cattle according to the abnormal score. If the abnormal score of the cattle is lower than the abnormal threshold, the cattle is healthy;

[0037] S412. If the abnormal score of the cattle is within the abnormal threshold range, the cattle is sub-healthy;

[0038] S413. If the abnormal score of the cattle is higher than the abnormal threshold, the cattle is abnormal.

[0039] According to another aspect of the present invention, there is also provided an intelligent analysis system for cattle health conditions based on feature recognition. The intelligent analysis system includes a data preprocessing module, a feature extraction module, a model construction module, and a health analysis module;

[0040] The data preprocessing module is used to obtain cattle health data based on pre-deployed cameras and sensors, and preprocess the cattle health data;

[0041] The feature extraction module is used to perform feature recognition on the preprocessed cattle health data by using the SIFT algorithm, confirm the cattle health feature elements, and construct a cattle health assessment feature set;

[0042] The model construction module is used to construct and train a cattle health analysis model based on the cattle health assessment feature set and the cattle health data by using the isolation forest algorithm, and determine the abnormal score;

[0043] The health analysis module is used to analyze the cattle health data by using the trained cattle health analysis model, and analyze and evaluate the health status of the cattle in combination with the abnormal score.

[0044] The beneficial effects of the present invention are as follows:

[0045] 1. The present invention realizes the automatic collection of cattle health data by deploying intelligent cameras and sensors, covering multi-modal data such as behavior, image, physiology, and environment. Combined with machine learning algorithms, the evaluation of the health status is automated, eliminating the subjective factors in manual observation. The automated evaluation improves efficiency and consistency, and is suitable for large-scale cattle herd health management.

[0046] 2. The present invention uses data cleaning and normalization technologies to convert data from different sources into a unified format, eliminating the format differences between data. By using the SIFT algorithm to extract key feature points in the image, and combining behavior, physiology, and environment data, a unified health assessment feature set is constructed, providing high-quality data input for subsequent health analysis.

[0047] 3. In the present invention, the Isolation Forest algorithm quickly separates normal and abnormal data through random partitioning, avoiding the limitations of traditional fixed-threshold methods. It can accurately detect outliers in data without label information, especially suitable for the detection of sub-healthy states. The abnormal scoring mechanism makes the health assessment more flexible, and can dynamically adjust the classification criteria for healthy, sub-healthy, and abnormal states according to historical data.

[0048] 4. This solution quickly classifies the health status by calculating the abnormal score in real time and setting a threshold, realizes the real-time monitoring of the cattle herd. The early warning mechanism marks the individuals with health abnormalities and generates records, facilitating timely intervention by the management personnel. The real-time monitoring greatly shortens the detection time of health abnormalities and reduces the impact of health problems on economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0050] Figure 1 is a flowchart of an intelligent analysis method for cattle health status based on feature recognition according to an embodiment of the present invention;

[0051] Figure 2 is a schematic block diagram of an intelligent analysis system for cattle health status based on feature recognition according to an embodiment of the present invention.

[0052] In the figure:

[0053] 1. Data preprocessing module; 2. Feature extraction module; 3. Model construction module; 4. Health analysis module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0054] To further illustrate the embodiments, the present invention provides drawings. These drawings are part of the disclosure of the present invention, mainly used to illustrate the embodiments, and can be combined with the relevant descriptions in the specification to explain the operating principle of the embodiments. With reference to these contents, those of ordinary skill in the art should be able to understand other possible implementation manners and the advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are usually used to represent similar components.

[0055] According to an embodiment of the present invention, an intelligent analysis method and system for cattle health status based on feature recognition are provided.

[0056] Now, the present invention will be further described in combination with the drawings and specific implementation manners, as Figure 1As shown, according to an embodiment of the present invention, an intelligent analysis method for the health condition of cows based on feature recognition is provided. The intelligent analysis method includes the following steps:

[0057] S1. Based on pre-deployed cameras and sensors, obtain cow health data and preprocess the cow health data.

[0058] In one embodiment, the step of obtaining cow health data based on pre-deployed cameras and sensors and preprocessing the cow health data includes the following steps:

[0059] S11. Through pre-deployed cameras and sensors, collect cow health data, where the cow health data includes daily behavior images and videos of cows and physiological data of cows;

[0060] S12. Perform data cleaning and normalization on the collected cow health data, and convert the processed data into a unified standard format.

[0061] It should be noted that video data collection is to obtain daily behavior images or video streams of cows through cameras, including postures such as standing, lying, and walking; sensor data collection is to collect physiological data of cows (such as body temperature, heart rate, rumination times, feed intake) in real time through sensors (such as intelligent collars or wearable devices); data such as environmental temperature, humidity, and air quality (such as ammonia concentration) are obtained through environmental sensors deployed in the cowshed.

[0062] In addition, the preprocessing of the data includes:

[0063] 1) Data cleaning, process outliers in the collected data (such as extreme values in sensor readings), and remove incomplete frames in the video (such as blurry or cow body occlusion).

[0064] 2) Time alignment, align the timestamps of data from different sources (video, sensor, environment) to ensure data synchronization.

[0065] 3) Image processing, remove the background from the video frames, segment the cow body region (ROI, Region of Interest), and perform image grayscaling and normalization processing to reduce the computational complexity.

[0066] 4) Data formatting, uniformly convert physiological, behavioral, and environmental data into a standard format for subsequent feature extraction and analysis.

[0067] S2. For the preprocessed cow health data, use the SIFT algorithm for feature recognition, confirm the cow health feature elements, and construct a cow health assessment feature set.

[0068] In one embodiment, for the preprocessed bovine health data, the SIFT algorithm is used for feature recognition to confirm the bovine health feature elements, and constructing a bovine health assessment feature set includes the following steps:

[0069] S21. Use the SIFT algorithm to extract key feature points from the preprocessed bovine health data.

[0070] In one embodiment, using the SIFT algorithm to extract key feature points from the preprocessed bovine health data includes the following steps:

[0071] S211. For the preprocessed bovine health data, use the SIFT algorithm to detect the key feature points of the images in the bovine health data, and through the construction of the Gaussian pyramid, locate the distribution of the key feature points;

[0072] S212. Based on the distribution of the key feature points, generate the feature vectors corresponding to each key feature point, and calculate the gradient distribution around the key feature points.

[0073] S22. Based on the bovine health data, combined with the extracted key feature points, construct a health assessment feature set.

[0074] It should be noted that when using the SIFT algorithm to extract key feature points from the preprocessed bovine health data, for the preprocessed images or video frames, the SIFT algorithm is used to detect the key points (Keypoints) in the images, including important parts such as the bovine body contour, limbs, and head. Through the construction of the Gaussian pyramid, the distribution of the key points at different scales and resolutions is located to ensure that the detection results are not affected by image scaling and illumination changes.

[0075] In addition, for each key point, generate its corresponding SIFT local descriptor, calculate the feature vector to describe the gradient distribution around the key point. The descriptor is used to capture changes in bovine behaviors (such as lying, standing, walking) and posture features, save the coordinates and distribution patterns of the key points and descriptors, provide input for subsequent analysis, align the SIFT feature points in the image with the physiological data (such as body temperature, heart rate, rumination times) collected by sensors, and integrate environmental data (such as temperature, humidity, ammonia concentration) with the bovine behavior features to form a spatio-temporal associated data set.

[0076] Construct a comprehensive health assessment feature set, including:

[0077] 1) Image features, such as the key point distribution pattern, local gradient direction, the number and change trend of key points.

[0078] 2) Behavior features, such as movement trajectories, feeding behaviors, and changes in resting time.

[0079] 3) Physiological features, such as body temperature fluctuations and heart rate patterns.

[0080] 4) Environmental characteristics, such as the association between temperature, humidity and health behaviors.

[0081] Simplify the feature set through a dimensionality reduction algorithm (such as PCA), and retain the most representative health factors.

[0082] S3. Based on the cattle health assessment feature set and cattle health data, use the Isolation Forest algorithm to construct and train a cattle health analysis model, and determine the anomaly score.

[0083] In one embodiment, the step of constructing and training a cattle health analysis model and determining the anomaly score based on the cattle health assessment feature set and cattle health data by using the Isolation Forest algorithm includes the following steps:

[0084] S31. Use the Isolation Forest algorithm, combined with the cattle health assessment feature set, to construct a cattle health analysis model and determine the anomaly score.

[0085] In one embodiment, the step of using the Isolation Forest algorithm, combined with the cattle health assessment feature set, to construct a cattle health analysis model and determine the anomaly score includes the following steps:

[0086] S311. Based on the Isolation Forest algorithm, formulate an anomaly score for the cattle health assessment feature set in combination with the cattle health data;

[0087] S312. Separate the outliers in the cattle health assessment feature set through random partitioning. If the data in the cattle health assessment feature set is lower than the anomaly score, it is an outlier;

[0088] S313. If the data in the cattle health assessment feature set reaches or exceeds the anomaly score, it is a normal value.

[0089] In one embodiment, the expression of the anomaly score is:

[0090] s(x,n) = 2 - c(n)E(h(x))

[0091] In the formula, s(x,n) represents the anomaly score, E(h(x)) represents the average separation depth of the data point x in the cattle health assessment feature set on all random trees, c(n) represents the regularization factor of the average separation depth of the random trees, and n represents the number of samples in the cattle health assessment feature set.

[0092] S32. Train and optimize the cattle health analysis model, and evaluate the performance of the cattle health analysis model.

[0093] It should be noted that the core idea of using the Isolation Forest algorithm - quickly separating outliers through random partitioning - is used to design a cattle health analysis model and define an Anomaly Score. The characteristic points of healthy cattle are more concentrated and have more partitions, while the characteristic points of abnormal cattle are more discrete and have fewer partitions, resulting in a higher score. In the input and output of the model, the input is the cattle health assessment feature set, including the integrated data of images, behaviors, physiology, and environment; the output is the anomaly score of each cattle, which is used to evaluate the health status; the score threshold is defined according to historical annotation data to divide the health status (such as healthy, sub-healthy, abnormal).

[0094] In addition, an unsupervised learning method is used to train the health analysis model using the Isolation Forest. During the training process, random Isolation Trees are generated to isolate data points and learn the feature distribution; partial annotated data is used to verify the model performance, including accuracy, which is the accuracy of the model in predicting the health status; recall rate, which is the proportion of the number of abnormal cattle detected by the model to the total number of abnormal cattle; and F1 score, which is used to evaluate the balanced performance of the model in anomaly detection; hyperparameters (such as the number of Isolation Trees and the depth of sample partitioning) are optimized to improve the detection ability of the model.

[0095] S4. Use the trained cattle health analysis model to analyze the cattle health data in combination with the anomaly score, and analyze and evaluate the cattle health status.

[0096] In one embodiment, the step of using the trained cattle health analysis model to analyze the cattle health data in combination with the anomaly score and analyze and evaluate the cattle health status includes the following steps:

[0097] S41. Calculate the anomaly score of the cattle in real-time based on the cattle health analysis model, set the anomaly threshold according to the anomaly score, and divide the cattle health status.

[0098] In one embodiment, the step of calculating the anomaly score of the cattle in real-time based on the cattle health analysis model, setting the anomaly threshold according to the anomaly score, and dividing the cattle health status includes the following steps:

[0099] S411. Calculate the anomaly score of the cattle in real-time based on the cattle health analysis model, divide the cattle health status according to the anomaly score. If the anomaly score of the cattle is lower than the anomaly threshold, the cattle is healthy;

[0100] S412. If the anomaly score of the cattle is within the anomaly threshold range, the cattle is sub-healthy;

[0101] S413. If the anomaly score of the cattle is higher than the anomaly threshold, the cattle is abnormal.

[0102] S42. Issue a warning for the cattle health status, record the cattle with health abnormalities, and report them.

[0103] It should be noted that when using the trained model to perform real-time analysis on health data to generate health scores, the real-time data input is to input the health data collected in real-time by the camera and sensors into the trained model; the preprocessing and SIFT feature extraction module is used to generate a real-time health assessment feature set and calculate the anomaly score for each cow in real-time.

[0104] If the anomaly score of a certain cow exceeds the threshold, a notification will be sent to the farm manager or veterinarian, including: the ear tag number of the cow, the anomaly score, the corresponding health problems, and possible reasons for the anomaly (such as too high body temperature, reduced activity), and then suggestions for measures will be given according to the actual situation:

[0105] 1) According to the type of anomaly, it is recommended to isolate, further examine, or adjust the feed and environment.

[0106] 2) For cows with abnormal feed intake, it is recommended to check the feed quality.

[0107] 3) For cows with isolated behavior, it is recommended to observe group interactions or check physiological data.

[0108] 4) Collect the feedback from the manager or veterinarian and compare the diagnosis results with the model predictions.

[0109] 5) Add the feedback data to the training set of the model and update the model regularly to improve the accuracy.

[0110] According to another aspect of the present invention, an intelligent analysis system for the health condition of cows based on feature recognition is also provided. As Figure 2 shown, the intelligent analysis system includes a data preprocessing module 1, a feature extraction module 2, a model construction module 3, and a health analysis module 4;

[0111] The data preprocessing module 1 is used to obtain cow health data based on pre-deployed cameras and sensors and perform preprocessing on the cow health data;

[0112] The feature extraction module 2 is used to perform feature recognition on the preprocessed cow health data using the SIFT algorithm, confirm the cow health feature elements, and construct a cow health assessment feature set;

[0113] The model construction module 3 is used to construct and train a cow health analysis model using the isolation forest algorithm based on the cow health assessment feature set and cow health data to determine the anomaly score;

[0114] The health analysis module 4 is used to analyze the cow health data using the trained cow health analysis model and analyze and evaluate the cow health status in combination with the anomaly score.

[0115] To facilitate the understanding of the above technical solution of the present invention, the working principle or operation mode of the present invention in the actual process will be described in detail below.

[0116] In summary, by means of the above technical solution of the present invention, the present invention realizes the automatic collection of cattle health data through the deployment of intelligent cameras and sensors, covering multi-modal data such as behavior, image, physiology, and environment. Combining machine learning algorithms, the evaluation of the health status is automated, eliminating the subjective factors in manual observation. The automated evaluation improves efficiency and consistency, and is suitable for large-scale cattle herd health management; the present invention uses data cleaning and normalization technologies to convert data from different sources into a unified format, eliminating the format differences between data. The SIFT algorithm is used to extract key feature points in the image, and combined with behavior, physiology, and environmental data, a unified health evaluation feature set is constructed to provide high-quality data input for subsequent health analysis; in the present invention, the isolation forest algorithm quickly separates normal and abnormal data through random partitioning, avoiding the limitations of traditional fixed threshold methods, and can accurately detect outliers in the data without label information, especially suitable for the detection of sub-healthy states. The abnormal scoring mechanism makes the health evaluation more flexible, and the division criteria of healthy, sub-healthy, and abnormal states can be dynamically adjusted according to historical data; this solution quickly divides the health status by calculating the abnormal score in real time and setting a threshold, realizes the real-time monitoring of the cattle herd, and the early warning mechanism marks the individuals with health abnormalities and generates records, facilitating the timely intervention of management personnel. The real-time monitoring greatly shortens the detection time of health abnormalities and reduces the impact of health problems on economic benefits.

[0117] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for intelligent analysis of cattle health based on feature recognition, characterized in that: The intelligent analysis method comprises the following steps: S1. Obtain cattle health data based on pre-deployed cameras and sensors, and pre-process the cattle health data; S2. Using SIFT algorithm to perform feature recognition on the preprocessed cattle health data, confirm the cattle health feature elements, and construct a cattle health assessment feature set; S3. Based on the cattle health assessment feature set and cattle health data, the cattle health analysis model is constructed and trained using the isolation forest algorithm to determine the abnormal score; S4. Use the trained cattle health analysis model and the abnormal score to analyze the cattle health data and analyze and evaluate the cattle health status.

2. The method for intelligent analysis of cattle health based on feature recognition according to claim 1, characterized in that: The method of obtaining cattle health data based on pre-deployed cameras and sensors and pre-processing the cattle health data includes the following steps: S11. Collecting cattle health data of cattle through pre-deployed cameras and sensors, wherein the cattle health data includes daily behavior images and videos of cattle and physiological data of cattle; S12. Clean and normalize the collected cattle health data, and convert the processed data into a unified standard format.

3. The method for intelligent analysis of cattle health status based on feature recognition according to claim 1, characterized in that: The method of using SIFT algorithm to perform feature recognition on the pre-processed cattle health data, confirming cattle health feature elements, and constructing a cattle health assessment feature set includes the following steps: S21, extract key feature points from the pre-processed cattle health data using SIFT algorithm; S22. Based on cattle health data and combined with the extracted key feature points, a health assessment feature set is constructed.

4. The method for intelligent analysis of cattle health based on feature recognition according to claim 3 is characterized in that: The method of extracting key feature points from the pre-processed cattle health data using the SIFT algorithm comprises the following steps: S211, using the SIFT algorithm to detect key feature points of the image in the preprocessed cattle health data, and constructing a Gaussian pyramid to locate the distribution of the key feature points; S212: Based on the distribution of key feature points, generate a feature vector corresponding to each key feature point, and calculate the gradient distribution around the key feature point.

5. The method for intelligent analysis of cattle health status based on feature recognition according to claim 1, characterized in that: The method of constructing and training a cattle health analysis model based on a cattle health assessment feature set and cattle health data using an isolation forest algorithm to determine an abnormal score includes the following steps: S31, using the isolation forest algorithm, combined with the cattle health assessment feature set, to build a cattle health analysis model and determine the abnormal score; S32. Train and optimize the cattle health analysis model, and evaluate the performance of the cattle health analysis model.

6. The method for intelligent analysis of cattle health conditions based on feature recognition according to claim 5, characterized in that: The method of using the isolation forest algorithm in combination with the cattle health assessment feature set to construct a cattle health analysis model and determine the abnormal score includes the following steps: S311, based on the isolation forest algorithm, develop anomaly scores for the cattle health assessment feature set in combination with cattle health data; S312, separating outliers in the cattle health assessment feature set by random partitioning, and if the data in the cattle health assessment feature set is lower than the abnormal score, it is an outlier; S313. If the data in the cattle health assessment feature set reaches or exceeds the abnormal score, it is considered normal.

7. The method for intelligent analysis of cattle health status based on feature recognition according to claim 6, characterized in that: The expression of the abnormality score is: s(x,n)=2-c(n)E(h(x)) Where s(x,n) represents the anomaly score, E(h(x)) represents the average separation depth of the data point x in the cattle health assessment feature set on all random trees, c(n) represents the regularization factor of the average separation depth of the random trees, and n represents the number of samples in the cattle health assessment feature set.

8. The method for intelligent analysis of cattle health status based on feature recognition according to claim 1, characterized in that: The method of using the trained cattle health analysis model and combining the abnormality score to analyze the cattle health data and analyze and evaluate the cattle health status includes the following steps: S41, calculating an abnormal score for the cattle in real time based on the cattle health analysis model, setting an abnormal threshold according to the abnormal score, and classifying the cattle health status; S42. Issue early warnings on the health status of cattle, record cattle with abnormal health conditions, and report them.

9. The method for intelligent analysis of cattle health conditions based on feature recognition according to claim 8, characterized in that: The method of calculating an abnormal score for a cattle in real time based on the cattle health analysis model, setting an abnormal threshold according to the abnormal score, and classifying the cattle health status comprises the following steps: S411, calculating an abnormality score for the cattle in real time based on the cattle health analysis model, and classifying the cattle health status according to the abnormality score. If the abnormality score of the cattle is lower than the abnormality threshold, the cattle is healthy; S412: If the abnormal score of the cattle is within the abnormal threshold range, the cattle is sub-healthy; S413. If the abnormality score of the cattle is higher than the abnormality threshold, the cattle is abnormal.

10. A cattle health condition intelligent analysis system based on feature recognition, used to implement a cattle health condition intelligent analysis method based on feature recognition as claimed in any one of claims 1 to 9, characterized in that: The intelligent analysis system includes a data preprocessing module, a feature extraction module, a model building module and a health analysis module; The data preprocessing module is used to obtain cattle health data based on pre-deployed cameras and sensors, and pre-process the cattle health data; The feature extraction module is used to perform feature recognition on the pre-processed cattle health data using the SIFT algorithm, confirm the cattle health feature elements, and construct a cattle health assessment feature set; The model building module is used to build and train a cattle health analysis model based on the cattle health assessment feature set and cattle health data using an isolation forest algorithm to determine an abnormal score; The health analysis module is used to analyze the health data of cattle by using the trained cattle health analysis model in combination with the abnormality score, and to analyze and evaluate the health status of cattle.

Images