A beef cattle health state evaluation method and system based on multi-source data fusion
By integrating triaxial acceleration, breeding management and environmental data, and combining machine learning algorithms, a health assessment method based on multi-source data fusion was developed to achieve real-time and accurate assessment of the health status of beef cattle. This solved the problems of inaccurate assessment and low efficiency in existing technologies and is suitable for large-scale intelligent breeding scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUIZHOU GUIGU AGRI CO LTD
- Filing Date
- 2026-03-06
- Publication Date
- 2026-06-09
AI Technical Summary
Existing beef cattle health assessment technologies rely on a single data source, resulting in inaccurate and inefficient assessments. They also lack a multi-source data fusion mechanism, making it difficult to adapt to the needs of large-scale and intelligent farming.
By integrating information on beef cattle triaxial acceleration, breeding management, environmental temperature and humidity, and individual identification, a multi-source data fusion model is established. A hierarchical fusion strategy is adopted to assess health status, and machine learning algorithms are combined to achieve real-time and accurate assessment.
It improves the accuracy and efficiency of beef cattle health status assessment, reduces reliance on manual labor, adapts to the needs of different farming scenarios, and enhances farming efficiency and product quality.
Smart Images
Figure CN122177438A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent beef cattle farming technology, specifically involving a method and system for assessing the health status of beef cattle based on multi-source data fusion. It is particularly suitable for real-time monitoring, risk warning, and status assessment of individual beef cattle in large-scale beef cattle farms, and achieves accurate judgment of the health status of beef cattle by integrating multi-dimensional monitoring data. Background Technology
[0002] Beef cattle farming is an important part of animal husbandry, and the health status of beef cattle directly affects farming efficiency, product quality, and the sustainable development of the industry. Currently, beef cattle health assessment mainly relies on manual observation by farmers, which has problems such as strong subjectivity, slow response, high rates of missed and incorrect judgments, and high labor intensity, making it difficult to adapt to the needs of large-scale and intelligent farming models.
[0003] With the development of intelligent farming technology, some farms have begun to introduce single-type monitoring equipment, such as using accelerometers to monitor the activity status of beef cattle, using temperature and humidity sensors to monitor the farming environment, or manually entering some farming management data. However, existing technologies have obvious shortcomings: First, the data sources are limited, and relying on only one type of data cannot comprehensively reflect the health status of beef cattle. For example, judging solely by activity levels can easily overlook the impact of environmental and farming management factors on the health of beef cattle, leading to insufficient assessment accuracy. Second, multi-source data lacks an effective fusion mechanism. Various types of data are independent of each other, failing to leverage data synergy and making it difficult to uncover hidden patterns in beef cattle health. Third, data collection methods are cumbersome, with some farming management data relying on manual entry, which is inefficient and prone to errors. At the same time, there is a lack of a unified data integration and analysis platform, making it impossible to achieve real-time assessment and early warning of health status.
[0004] Therefore, there is an urgent need for a method and system for assessing the health status of beef cattle that can integrate multi-source monitoring data, achieve efficient data fusion and accurate analysis, and be adapted to large-scale intelligent farming scenarios, in order to solve the problems of inaccurate assessment, low efficiency and insufficient intelligence in existing technologies. Summary of the Invention
[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a method and system for assessing the health status of beef cattle based on multi-source data fusion. By integrating beef cattle triaxial acceleration data, breeding management data, environmental temperature and humidity data, and individual identity information, a multi-source data fusion health assessment model is established to achieve real-time, accurate, and comprehensive assessment of the health status of beef cattle, reduce reliance on manual labor, and improve the level of intelligent breeding and breeding efficiency.
[0006] A method for assessing the health status of beef cattle based on multi-source data fusion. The method includes the following steps:
[0007] Step 1: Multi-source data acquisition. Multi-source data of individual beef cattle is acquired simultaneously using multiple types of acquisition devices. This multi-source data includes triaxial acceleration data, livestock management data, environmental temperature and humidity data, and individual identification information. The data acquisition methods for each type are as follows: Individual identification information: Each individual beef cattle is individually linked to it to associate various data, enabling accurate individual identification and health assessment; Three-axis acceleration data: collected by a three-axis acceleration sensor built into a smart neck collar worn by beef cattle, used to reflect the activity status of beef cattle (such as feeding, standing, resting, walking, etc.). The collection frequency can be set from 1 to 100Hz as needed and transmitted to the data processing terminal in real time. Environmental temperature and humidity data: collected by temperature and humidity sensors deployed in the breeding area, covering all areas where beef cattle are active, to monitor the temperature and humidity parameters of the breeding environment in real time and transmit them to the data processing terminal; Livestock management data is acquired through multi-device collaboration. Specifically, this includes: basic information (breed, age, manually entered or imported via a livestock farming app when cattle enter the pen), weight data (collected through a stress-free automatic weighing system to monitor and upload cattle weight in real time without stress, avoiding any impact on cattle health during the weighing process), location data (collected through the BDS positioning module built into the smart collar to obtain real-time spatial location information and determine if the cattle's activity range is abnormal), feeding-related data (daily feeding time and frequency collected via the smart collar, feeding methods and feed formulas entered via the smart collar-assisted livestock farming app), and reproductive and disease prevention data (reproduction records, disease prevention records, and treatment records entered via the smart collar-assisted livestock farming app; reproduction records can be further verified by combining activity data collected by the smart collar). All types of collected data are linked to individual identification information to ensure data relevance and traceability.
[0008] Step 2: Data Preprocessing. The multi-source data collected in Step 1 is preprocessed to eliminate noise, missing values, and outliers, and to standardize the data format, laying the foundation for subsequent data fusion. This includes: Outlier handling: Use the 3σ principle or box plot method to identify and remove outliers (such as extreme values caused by sensor failure or abnormal data caused by data transmission errors) from triaxial acceleration data, weight data, and environmental temperature and humidity data. Missing value handling: For missing values in aquaculture management data and triaxial acceleration data, interpolation methods (linear interpolation, polynomial interpolation) or mean imputation methods are used to supplement them to ensure data integrity; Data standardization: Convert all types of data to the same order of magnitude (e.g., using the min-max standardization method) to eliminate the influence of dimensions. Triaxial acceleration data needs to be filtered (e.g., Kalman filtering) to remove motion noise. Data formatting: Convert all preprocessed data into a unified timestamp format, classify and store it according to individual identity information, and form a personalized data file for each beef cattle.
[0009] Step 3: Multi-source data fusion adopts a hierarchical fusion strategy, dividing the preprocessed multi-source data into three levels: data layer, feature layer, and decision layer for fusion, as detailed below: Data layer fusion: Individual identity information is associated and aligned with triaxial acceleration data, aquaculture management data, and environmental temperature and humidity data. Using timestamps and individual identities as indexes, a unified multi-source dataset is constructed to ensure the spatiotemporal consistency of various types of data. Feature layer fusion: Health-related features of various data are extracted, and feature concatenation and weighted fusion algorithms are used to fuse various features, assigning different weights to different features (weights can be dynamically adjusted according to the breeding scenario and historical data, such as triaxial acceleration features and weight change rate features having higher weights than environmental temperature and humidity features), to obtain a fused feature vector for beef cattle health assessment. Triaxial acceleration data features extract features such as activity frequency, activity intensity, feeding duration, and resting duration, reflecting the physiological activity state of beef cattle; environmental temperature and humidity data features extract features such as mean temperature and humidity, fluctuation range, and duration of extreme values, reflecting the impact of the breeding environment on beef cattle health; breeding management data features extract features such as weight change rate, feeding patterns (daily feeding frequency, duration of each feeding), disease prevention compliance, and rationality of the breeding cycle, reflecting the breeding management status of beef cattle; Decision-level fusion: The fused feature vector is input into a pre-set health assessment model. Combined with historical health data and labeled samples, the health status of beef cattle is classified and assessed using machine learning algorithms (such as random forest, support vector machine, and deep learning model), and the health assessment results are output.
[0010] Step 4: Health Status Assessment and Result Output. Based on the assessment results integrated by the decision-making level, the health status of beef cattle is divided into three levels: normal, sub-healthy, and diseased. Normal level: The fused feature vectors conform to the characteristic range of healthy beef cattle, and the activity, feeding, weight change and environmental adaptability of the beef cattle are all at normal levels; Sub-health level: The fused feature vector shows slight abnormalities (such as slightly lower than normal activity level, abnormal weight change rate, and large fluctuations in environmental temperature and humidity), suggesting that farmers should take targeted interventions (such as adjusting the feeding formula and improving the breeding environment). Disease severity: The fused feature vector shows obvious abnormalities (such as a sudden drop in activity, cessation of feed intake, abnormal weight loss, missing vaccination records and abnormal activity characteristics), indicating that the beef cattle may have health problems and need to be investigated and treated in time. The assessment results are linked with individual identity information, various raw data, and integrated data, and output to the breeding APP and data monitoring platform in real time. At the same time, early warnings are issued for sub-healthy and diseased beef cattle, and the reasons for the warnings are clearly stated (such as "abnormal activity level" or "rapid weight loss"), so as to facilitate precise handling by farmers.
[0011] Step 5: Model Optimization and Update. Regularly collect health feedback data of beef cattle (such as treatment effects and changes in health status after intervention) and new multi-source data to train and optimize the health assessment model. Dynamically adjust feature weights and assessment thresholds to improve the accuracy and adaptability of the assessment model and adapt it to the health assessment needs of different breeds and different breeding scenarios.
[0012] A beef cattle health status assessment system based on multi-source data fusion is disclosed. This system implements the aforementioned beef cattle health status assessment method and includes a data acquisition module, a data preprocessing module, a multi-source data fusion module, a health assessment module, a data storage module, a display and early warning module, and an individual identification module. These modules work collaboratively, and their specific structure is as follows:
[0013] The individual identification module is used to store and identify the individual identity information of beef cattle, realize the one-to-one binding of individual identity with various types of collected data, and ensure the relevance and traceability of data. It can be implemented by using RFID tags or smart neck rings with built-in identity identification modules.
[0014] The data acquisition module includes multiple types of acquisition units for synchronously collecting multi-source data from beef cattle. Specifically, these include: an acceleration acquisition unit (a three-axis accelerometer integrated into the smart neck collar for collecting triaxial acceleration data from beef cattle and transmitting it to the data preprocessing module); an environmental acquisition unit (temperature and humidity sensors deployed in the breeding area for collecting environmental temperature and humidity data and uploading it to the data preprocessing module in real time); and a breeding management data acquisition unit (including a breeding APP, a stress-free automatic weighing system, and the smart neck collar; wherein the breeding APP is used to input or import information such as beef cattle breed, age, breeding records, feeding methods, feeding formulas, disease prevention records, and treatment records; the stress-free automatic weighing system is used to collect beef cattle weight data; the smart neck collar's BDS positioning module is used to collect beef cattle location data; and the smart neck collar's built-in acquisition unit is used to collect the daily feeding time and frequency of beef cattle).
[0015] The data preprocessing module is connected to the data acquisition module and the individual identification module. It is used to receive various types of collected data and individual identification information, perform the data preprocessing operations in step 2 (outlier handling, missing value handling, data standardization, data formatting), and output the preprocessed standardized multi-source dataset.
[0016] The multi-source data fusion module is connected to the data preprocessing module. It adopts a hierarchical fusion strategy, performs the data layer, feature layer, and decision layer fusion operations in step 3, extracts the health correlation features of various types of data, and outputs the fused feature vector.
[0017] The health assessment module is connected to the multi-source data fusion module, with a built-in preset health assessment model (machine learning algorithm model). It receives the fused feature vector, performs the health status classification assessment in step 4, and outputs the health status assessment results of beef cattle (normal, sub-healthy, diseased) and early warning information.
[0018] The data storage module is connected to the data preprocessing module, the multi-source data fusion module, and the health assessment module. It is used to store raw collected data, preprocessed data, fused feature vectors, health assessment results, historical health data, assessment model parameters, etc., and supports data query, traceability, and updating.
[0019] The display and early warning module is connected to the health assessment module and data storage module, including the livestock APP display terminal and the data monitoring platform. It is used to display the individual identity information of beef cattle, various data, and health assessment results in real time, issue early warning prompts for sub-healthy or sick beef cattle, push the reasons for the early warning and treatment suggestions, and also support livestock farmers to view historical data and assessment reports.
[0020] Furthermore, the smart neck collar communicates with the livestock farming app and data processing terminal via a wireless communication module (such as 5G, 2.4G, NB-IoT, CATI) to ensure real-time data transmission. The stress-free automatic weighing system is deployed in the cattle activity channel to automatically collect the weight of cattle as they move around, without the need for manual intervention.
[0021] Furthermore, the health assessment module supports both manual and automatic model updates. Automatic updates can collect new data according to a preset cycle (such as weekly or monthly) to complete model training optimization.
[0022] Beneficial Effects Compared with the prior art, the present invention has the following significant beneficial effects: 1. Comprehensive data sources and high assessment accuracy: It integrates multi-source data on beef cattle triaxial acceleration, breeding management, environmental temperature and humidity, and individual identity, covering multiple dimensions such as beef cattle physiological activities, breeding management, and environmental adaptation. It overcomes the limitations of single data assessment, and significantly improves the accuracy of beef cattle health status assessment and reduces the rate of missed and false judgments by mining data synergistic correlations through multi-source data fusion. 2. High degree of intelligence and reduced reliance on manual labor: Data is collected collaboratively by multiple devices. Most of the breeding management data is automatically collected or assisted in the input through smart neck collars, stress-free weighing systems, and APP, reducing the workload and error of manual input; it realizes real-time health status assessment and automatic early warning, eliminating the need for farmers to observe in real time, greatly reducing labor intensity and adapting to large-scale beef cattle breeding scenarios; 3. Strong data correlation and good traceability: All collected data are linked one-to-one with the individual identity information of beef cattle, constructing personalized beef cattle data profiles, realizing full-process traceability from data collection, integration, evaluation to early warning and disposal, which facilitates farmers to investigate the causes of health problems and optimize breeding management strategies; 4. Highly adaptable and dynamically optimizable: The evaluation model supports dynamic optimization based on historical data and farming scenarios. Feature weights can be flexibly adjusted to adapt to the health assessment needs of beef cattle of different breeds, ages, and feeding methods. At the same time, the data collection types and evaluation dimensions can be expanded according to the upgrading of farming technology. 5. Timely early warning, improving breeding efficiency: It can identify the sub-healthy state of beef cattle in advance and issue targeted warnings, which makes it easier for farmers to intervene in time and prevent the sub-healthy state from developing into a diseased state; at the same time, it can quickly identify sick beef cattle, help timely treatment, reduce the morbidity and mortality of beef cattle, and improve breeding efficiency and product quality. Attached Figure Description
[0023] Figure 1This is an overall architecture diagram of the beef cattle health status assessment system based on multi-source data fusion according to the present invention. The diagram clearly shows the composition of each functional module of the assessment system and the connection relationship between the modules. The individual identification module is connected to the data acquisition module and is used to bind individual identity information to various types of collected data. The data acquisition module includes three sub-units: an acceleration acquisition unit, an environmental acquisition unit, and a breeding management data acquisition unit. It is responsible for synchronously collecting multi-source data and transmitting it to the data preprocessing module. The data preprocessing module, the multi-source data fusion module, and the health assessment module are connected in series to realize the process connection of data preprocessing, hierarchical fusion, and health assessment. The health assessment module is connected to the display and early warning module, the data storage module, and the model optimization module. It can output the assessment results and early warning information to the display and early warning module, store various types of data and model parameters to the data storage module, and receive optimization parameters from the model optimization module to complete the model update. The data storage module can also feed back historical data to the health assessment module to help improve the accuracy of the assessment. All modules work together to realize all the functions of the system of the present invention.
[0024] Figure 2 This is a flowchart illustrating the steps of the beef cattle health status assessment method based on multi-source data fusion according to the present invention. The flowchart visually presents the complete execution flow of the assessment method, starting with "Start," and sequentially executing steps 1 (multi-source data acquisition), 2 (data preprocessing), 3 (multi-source data fusion), 4 (health status assessment and result output), and 5 (model optimization and update), ending with "End, cyclic execution." This clearly reflects the order of each step and the closed-loop process. Step 1 involves acquiring multi-source data on individual beef cattle identity, triaxial acceleration, breeding management, and environmental temperature and humidity. Step 2 preprocesses the collected data to eliminate noise and standardize the format. Step 3 obtains a fusion feature vector through a hierarchical fusion strategy. Step 4 outputs a three-level health assessment result and early warning information based on the fusion feature vector. Step 5 optimizes model parameters through feedback data to ensure the accuracy and adaptability of the assessment method. This flowchart fully corresponds to the core technical solution of the assessment method of the present invention and is the accompanying diagram in the abstract, used to summarize the core inventive points of the present invention. Detailed Implementation
[0025] The present invention will be further described in detail below with reference to specific embodiments, so that those skilled in the art can understand it.
[0026] Example 1: A method for assessing the health status of beef cattle based on multi-source data fusion
[0027] This embodiment is applied to a large-scale beef cattle farm with a breeding scale of 500 beef cattle. Each beef cattle wears a smart neck collar (with a built-in three-axis accelerometer, BDS positioning module, and wireless communication module). Temperature and humidity sensors are deployed in the breeding area (one per 500 square meters). A stress-free automatic weighing system is provided (deployed at the entrance of the beef cattle feeding passage). Farmers use a breeding APP to input and manage data.
[0028] Step 1: Multi-source data acquisition assigns a unique individual identification to each beef cattle, binding it to a smart collar and RFID tag; the smart collar's built-in triaxial accelerometer collects triaxial acceleration data at a frequency of 5Hz, the BDS positioning module collects the cattle's location data every 60 minutes, and simultaneously collects the daily feeding time and frequency; temperature and humidity sensors collect real-time temperature and humidity data of the breeding environment (collected every 5 minutes); a stress-free automatic weighing system collects the cattle's weight data (collected every 7 days); when the cattle enter the pen, the farmers enter breed and age information through the breeding APP, and daily records through the APP include breeding records, feeding methods (free-range + supplementary feeding), feeding formula (60% corn, 20% soybean meal, 20% straw), disease prevention records (foot-and-mouth disease prevention once a month), and treatment records; all data is transmitted in real-time to the data processing terminal via a wireless communication module and associated with the individual identification.
[0029] Step 2: Data preprocessing uses the 3σ principle to remove outliers in the triaxial acceleration data (such as acceleration extremes caused by sensor failure), and linear interpolation to supplement missing values in the weight data (such as missing values caused by some cattle not passing through the weighing channel); the min-max standardization method is used to standardize the triaxial acceleration data, weight data, and environmental temperature and humidity data to the [0,1] interval, and Kalman filtering is used to remove motion noise from the triaxial acceleration data; all data are converted into a unified timestamp format, classified and stored according to individual identification, forming a personalized data file for each cattle.
[0030] Step 3: Multi-source data fusion. Data layer fusion: Using timestamps and individual identity identifiers as indexes, align and link individual identity information with triaxial acceleration data, breeding management data, and environmental temperature and humidity data to construct a unified multi-source dataset. Feature layer fusion: Extract the mean activity intensity, feeding duration, and resting duration features from the triaxial acceleration data; extract the daily average and fluctuation amplitude features from the environmental temperature and humidity data; extract the weight change rate (calculated weekly), daily feeding frequency (twice / day), and disease prevention compliance (no missed prevention) features from the breeding management data. Use a weighted fusion algorithm to assign a weight of 0.4 to the triaxial acceleration feature, 0.2 to the weight change rate feature, 0.15 to the environmental temperature and humidity feature, and 0.25 to the other breeding management data features to obtain a fused feature vector. Decision layer fusion: Input the fused feature vector into a random forest evaluation model (trained based on the health data of 1000 historical beef cattle) to perform health status classification and evaluation.
[0031] Step 4: Health Status Assessment and Result Output. The assessment model outputs that 3 beef cattle are in a sub-healthy state, with warning reasons for "activity level 20% lower than normal", "weight change rate -5% (abnormal decrease)" and "environmental humidity higher than 80% for 3 consecutive days". The remaining 497 beef cattle are in a normal state. The warning module pushes the assessment results and warning information to the breeding APP and monitoring platform. According to the warning reasons, the breeders intervene in the sub-healthy beef cattle: adjust feeding time, increase activity space, improve ventilation in the breeding environment, and reduce environmental humidity.
[0032] Step 5: Model Optimization and Update One month later, we collected post-intervention health data (normal activity level, stable weight, and good environmental adaptability) and newly added multi-source data from three sub-healthy beef cattle. We trained and optimized the random forest evaluation model, adjusted the feature weights (increasing the environmental humidity feature weight to 0.2), and improved the model's accuracy in identifying sub-health states caused by environmental factors.
[0033] Example 2: A Beef Cattle Health Status Assessment System Based on Multi-Source Data Fusion
[0034] The system in this embodiment corresponds to the method in Embodiment 1, and includes an individual identification module, a data acquisition module, a data preprocessing module, a multi-source data fusion module, a health assessment module, a data storage module, and a display and early warning module.
[0035] The individual identification module uses RFID tags, one for each beef cattle, which are bound to the built-in identification mark in the smart neck collar for accurate individual identification. The data acquisition module includes a smart neck collar (triaxial accelerometer, BDS positioning module), temperature and humidity sensors, a stress-free automatic weighing system, and a livestock farming APP. Each acquisition unit collects various types of data simultaneously and transmits them wirelessly. The data preprocessing module uses an industrial computer to handle outliers and missing values and perform standardization operations. The multi-source data fusion module integrates a hierarchical fusion algorithm to achieve fusion of the data layer, feature layer, and decision layer. The health assessment module has a built-in random forest assessment model deployed on a cloud server and supports automatic updates. The data storage module uses a cloud database to store various types of data and assessment results, supporting data query and traceability. The display and early warning module includes a livestock farming APP display terminal and a monitoring platform. Farmers can view real-time assessment results and early warning information through the APP, and the monitoring platform can realize batch monitoring of beef cattle health status and data statistical analysis.
[0036] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for assessing the health status of beef cattle based on multi-source data fusion, characterized in that, Includes the following steps: Step 1: Multi-source data acquisition. Multi-source data for individual beef cattle is simultaneously acquired using various types of acquisition devices. This multi-source data includes individual identification information, triaxial acceleration data, breeding management data, and environmental temperature and humidity data. The triaxial acceleration data is acquired through a smart collar worn around the cattle's neck, which incorporates a triaxial acceleration sensor. The breeding management data includes breed, age, weight, location, breeding records, feeding methods, feeding formula, daily feeding time and frequency, disease prevention records, and treatment records. Breed and age are entered or imported via a breeding app; weight is acquired through a stress-free automatic weighing system; location and daily feeding time and frequency are acquired through the smart collar; and breeding records, feeding methods, feeding formula, disease prevention records, and treatment records are entered into the breeding app with the assistance of the smart collar. The environmental temperature and humidity data is acquired through temperature and humidity sensors deployed in the breeding area. All data is linked to individual identification information. Step 2: Data preprocessing. The multi-source data collected in Step 1 is processed for outlier handling, missing value handling, data standardization and data formatting to eliminate data noise, missing values and outliers, unify data format and magnitude, and form a standardized multi-source dataset. Step 3: Multi-source data fusion, employing a hierarchical fusion strategy, sequentially performing data layer fusion, feature layer fusion, and decision layer fusion; the data layer fusion uses timestamps and individual identities as indexes to align and associate various types of data, constructing a unified multi-source dataset; the feature layer fusion extracts health-related features from various types of data, and uses feature concatenation and weighted fusion algorithms to fuse these features, obtaining a fused feature vector; the decision layer fusion inputs the fused feature vector into a preset health assessment model, combining it with historical health data and labeled samples for analysis; Step 4: Health status assessment and result output. Based on the fusion analysis results of the decision-making level, the health status of beef cattle is divided into three levels: normal, sub-healthy, and diseased. The assessment results, individual identity information and early warning information are output in real time, and targeted early warnings are issued for sub-healthy and diseased beef cattle. Step 5: Model optimization and update. Regularly collect beef cattle health feedback data and new multi-source data to train and optimize the preset health assessment model, and dynamically adjust feature weights and assessment thresholds.
2. The method according to claim 1, characterized in that, In step 1, the triaxial accelerometer has a sampling frequency of 1-100Hz, the smart neck collar has a built-in BDS positioning module for collecting cattle position data, and the stress-free automatic weighing system is deployed in the cattle's activity channel to realize automatic weight collection when cattle move autonomously.
3. The method according to claim 1, characterized in that, In step 2, outlier handling uses the 3σ principle or box plot method; missing value handling uses interpolation or mean imputation method; data standardization uses the min-max standardization method; the triaxial acceleration data also needs to be processed by Kalman filtering to remove motion noise; data formatting converts all preprocessed data into a unified timestamp format and stores them according to individual identity information.
4. The method according to claim 1, characterized in that, In step 3, the health-related features extracted by the feature layer fusion include: activity frequency, activity intensity, feeding duration, and resting duration features of triaxial acceleration data; mean temperature and humidity, fluctuation range, and extreme value duration features of environmental temperature and humidity data; and weight change rate, feeding pattern, disease prevention compliance, and reproductive cycle rationality features of aquaculture management data. The feature weights of the weighted fusion algorithm can be dynamically adjusted according to the aquaculture scenario and historical data.
5. The method according to claim 1, characterized in that, In step 3, the preset health assessment model uses machine learning algorithms, including random forest, support vector machine or deep learning model.
6. The method according to claim 1, characterized in that, In step 4, the normal level is defined as the fused feature vector conforming to the characteristics of a healthy beef cattle, with normal activity, feeding, weight changes, and environmental adaptability; the sub-healthy level is defined as the fused feature vector showing slight abnormalities, prompting farmers to take targeted interventions; and the disease level is defined as the fused feature vector showing significant abnormalities, indicating that the beef cattle may have health problems and need to be investigated and treated in a timely manner.
7. A beef cattle health status assessment system based on multi-source data fusion, characterized in that, The method for implementing any one of claims 1-6 includes an individual identification module, a data acquisition module, a data preprocessing module, a multi-source data fusion module, a health assessment module, a data storage module, and a display and early warning module; The individual identification module is used to store and identify the individual identity information of beef cattle, and to bind the individual identity with various types of collected data. It is implemented by using an RFID tag or a smart neck ring with a built-in identity identification module. The data acquisition module includes an acceleration acquisition unit, an environmental acquisition unit, and a breeding management data acquisition unit, used to simultaneously acquire and transmit multi-source data; the acceleration acquisition unit is a triaxial acceleration sensor built into a smart collar; the environmental acquisition unit is a temperature and humidity sensor; the breeding management data acquisition unit includes a breeding APP, a stress-free automatic weighing system, and a smart collar; The data preprocessing module is connected to the data acquisition module and the individual identification module, and is used to perform the preprocessing operation in step 2 and output a standardized multi-source dataset. The multi-source data fusion module is connected to the data preprocessing module and is used to perform the hierarchical fusion operation in step 3 and output the fusion feature vector. The health assessment module is connected to the multi-source data fusion module and has a built-in preset health assessment model. It is used to perform the health status assessment in step 4 and output the assessment results and early warning information. The data storage module is connected to the data preprocessing module, the multi-source data fusion module, and the health assessment module. It is used to store various types of data, assessment results, and model parameters, and supports data query, traceability, and updating. The display and early warning module is connected to the health assessment module and the data storage module, and includes a breeding APP display terminal and a data monitoring platform, which are used to display various information and issue early warning prompts.
8. The system according to claim 7, characterized in that, The smart collar communicates with the aquaculture app and data processing terminal via a wireless communication module, which can be a 5G, 2.4G, NB-IoT, or CAT1 module.
9. The system according to claim 7, characterized in that, The health assessment module supports both manual and automatic model updates. Automatic updates collect new data according to a preset cycle to complete model training and optimization.