Bee breeding health state analysis and optimization management method based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a bee breeding health state analysis and optimization management method based on artificial intelligence. The method comprises the following steps: analyzing bee area multi-dimensional sensing characteristic data through target bee area multi-source monitoring sensing data; analyzing bee behavior heterogeneous event combination feature data based on the bee area multi-dimensional perception feature data; analyzing bee breeding health state evaluation data through the bee behavior heterogeneous event combination feature data; based on the multi-dimensional perception characteristic data of the bee area and the evaluation data of the health state of bee breeding, related characteristic analysis of health states and environment indexes under different breeding behaviors is carried out, and bee breeding health state-environment related characteristic data is generated; and designing a bee optimization breeding intelligent management strategy through the bee breeding health state-environment related characteristic data. According to the invention, efficient bee breeding health state analysis and intelligent breeding optimization management are realized.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based method for analyzing and optimizing the health status of beekeeping. Background Technology

[0002] Currently, with the upgrading and development of modern agricultural industries, beekeeping, as an important component of ecological agriculture, is experiencing continuous improvement in its industrial scale and intensification. The health status of bees directly determines the survival efficiency of the bee colony, the quality of product output, and the pollination service capacity. The close relationship between bee colony health and the breeding environment and bee behavior characteristics makes precise health monitoring and scientific management optimization the core keys to improving beekeeping efficiency. In actual beekeeping scenarios, different regions, seasons, and breeding models all significantly impact bee colony health due to pest and disease threats, environmental stresses, and differences in beekeeping behaviors. However, existing beekeeping health status analysis and optimization management technologies suffer from fragmented monitoring methods and a lack of systematic integration of multi-source sensing data, making it impossible to achieve collaborative analysis of bee behavioral characteristics and breeding environment parameters. Analysis of bee behavior remains at a simple superficial level, failing to delve into the combined characteristics and evolutionary patterns of behavioral events, thus hindering the accurate identification of early health anomalies. Furthermore, a quantitative correlation model between bee health status and environmental factors and breeding behavior has not been established, resulting in a lack of scientific basis for health assessments and an inability to adapt to the personalized needs of different bee colonies. Finally, the lack of intelligent data analysis and strategy generation mechanisms makes it difficult to predict health risks and dynamically optimize breeding management. Summary of the Invention

[0003] Based on this, the present invention provides an artificial intelligence-based method for analyzing and optimizing the health status of beekeeping, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an artificial intelligence-based method for analyzing and optimizing the health status of beekeeping includes the following steps:

[0005] Step S1: Use multi-source sensors to perform multi-source monitoring and sensing processing on the target bee area to generate multi-source monitoring and sensing data of the target bee area; use the multi-source monitoring and sensing data of the target bee area to perform multi-dimensional hierarchical characteristic classification processing on the target bee area sensing and monitoring to generate multi-dimensional sensing characteristic data of the bee area.

[0006] Step S2: Perform heterogeneous event combination feature analysis on the multidimensional perception characteristic data of bee regions to generate heterogeneous event combination feature data of bee behavior;

[0007] Step S3: Perform beekeeping health status assessment processing on the combined feature data of heterogeneous bee behavior events to generate beekeeping health status assessment data;

[0008] Step S4: Based on the multidimensional perception characteristic data of bee areas and the health status assessment data of bee farming, conduct relevant feature analysis on the health status and environmental indicators under different farming behaviors, and generate bee farming health status-environment related feature data.

[0009] Step S5: Design intelligent management strategies for optimized beekeeping by using beekeeping health status and environmental-related characteristic data.

[0010] Furthermore, the multi-source monitoring and sensing data of the target bee area mentioned in step S1 includes bee behavior sensing data and bee area environmental data. The bee behavior sensing data includes bee flight image data, bee monitoring vibration signals, and bee monitoring voiceprint signals. The bee area environmental data includes area environmental temperature and humidity data and area environmental gas composition data.

[0011] Furthermore, step S1 includes the following steps:

[0012] Step S11: Design multi-source sensor topology configuration data;

[0013] Step S12: Use the multi-source sensor topology configuration data to perform multi-source sensor topology configuration operation on the multi-source sensor, and use the configured multi-source sensor to perform multi-source bee monitoring and sensing processing on the target bee area to generate multi-source monitoring and sensing data of the target bee area.

[0014] Step S13: Perform data consistency verification and adjustment processing on the multi-source monitoring and sensing data of the target bee area to generate standard multi-source monitoring and sensing data of the target bee area;

[0015] Step S14: Perform time-series and spatial calibration processing on the multi-source monitoring and sensing data of the standard target bee area to generate spatiotemporal target bee area multi-source monitoring and sensing data;

[0016] Step S15: Perform multi-dimensional hierarchical characteristic segmentation processing on the spatiotemporal target bee region multi-source monitoring and sensing data to generate bee region multi-dimensional sensing characteristic data.

[0017] Furthermore, step S11 includes the following steps:

[0018] Analyze bee activity correlation data in the target bee area to generate bee activity correlation data, wherein the bee activity correlation data includes beehive structure data and bee activity range data.

[0019] Based on the correlation data of bee activity, the distribution characteristics of bee activity are analyzed to generate bee activity distribution characteristic data, and multi-source sensor topology configuration data is designed based on the bee activity distribution characteristic data.

[0020] Furthermore, step S2 includes the following steps:

[0021] Step S21: Extract bee behavior subdomain modalities based on bee region multidimensional perception characteristic data to generate bee behavior subdomain modal data;

[0022] Step S22: Perform multimodal clustering feature analysis on the bee behavior subdomain modal data to generate multimodal clustering feature data on bee behavior, and perform bee behavior pattern analysis based on the multimodal clustering feature data on bee behavior to generate bee behavior pattern data;

[0023] Step S23: Perform bee behavior event analysis on the bee behavior subdomain modal data to generate bee behavior event data;

[0024] Step S24: Perform event combination difference analysis on bee behavior event data using bee behavior pattern data to generate bee behavior event combination difference data;

[0025] Step S25: Perform heterogeneous event combination feature analysis on bee behavior based on the data of differences in bee behavior event combinations, and generate feature data of heterogeneous event combinations of bee behavior.

[0026] Furthermore, step S23 includes the following steps:

[0027] Step S231: Perform multimodal temporal feature analysis of bee behavior based on the bee behavior subdomain modal data, generate multimodal temporal feature data of bee behavior, and design a dynamic temporal window for bee behavior events based on the multimodal temporal feature data of bee behavior;

[0028] Step S232: Perform bee behavior modality spatial feature analysis on the bee behavior subdomain modal data to generate bee behavior modality spatial feature data;

[0029] Step S233: Detect bee behavior events using the dynamic temporal window of bee behavior events to generate bee behavior event data.

[0030] Furthermore, step S3 includes the following steps:

[0031] Step S31: Perform a correlation feature analysis on the health status of beekeeping based on the combined feature data of heterogeneous events in bee behavior, and generate correlation feature data on the health status of beekeeping.

[0032] Step S32: Analyze the basic activity characteristics and operational behavior characteristics of bees based on the correlation characteristics data of bee health status;

[0033] Step S33: Analyze the evolutionary trend of bee behavior based on the basic activity characteristics data and operational behavior characteristics data of bees, and generate bee behavior evolutionary trend data;

[0034] Step S34: Obtain historical bee health status assessment data;

[0035] Step S35: Design a multi-level discrimination matrix relationship for bee breeding health status based on historical bee health status assessment data, bee operational behavior data, and bee behavior evolution trend data, and generate a multi-level discrimination matrix for bee breeding health status.

[0036] Step S36: Based on the multi-level discrimination matrix of bee breeding health status, perform bee breeding health status assessment processing on the combination feature data of heterogeneous events of bee behavior to generate bee breeding health status assessment data.

[0037] Furthermore, step S4 includes the following steps:

[0038] Step S41: Analyze the environmental distribution characteristics of bees based on the multidimensional perception characteristic data of bee regions, and generate bee environmental distribution characteristic data;

[0039] Step S42: Perform health status and environmental correlation processing on the beekeeping health status assessment data and bee environmental distribution characteristic data to generate beekeeping health status-environment correlation data;

[0040] Step S43: Based on the correlation data between beekeeping health status and environment, conduct health status intervention feature analysis on beekeeping behavior and environmental indicators to generate health status intervention feature data;

[0041] Step S44: Perform data segmentation processing on the health status interventionable feature data to conditionalize breeding behavior and the differences in environmental indicators, and generate breeding behavior conditionalization-environmental difference segmented data;

[0042] Step S45: Based on the conditionalized breeding behavior-environmental difference segmented data and the bee breeding health status-environmental correlation data, conduct an environmental impact characteristic analysis of breeding health status and generate environmental impact characteristic data of breeding health status.

[0043] Step S46: Based on the environmental impact characteristic data of beekeeping health status, conduct relevant characteristic analysis of health status and environmental indicators under different beekeeping behaviors to generate beekeeping health status-environment related characteristic data.

[0044] Furthermore, step S45 includes the following steps:

[0045] By analyzing the gradient changes in beekeeping health status based on environmental indicators through segmented data of beekeeping behavior conditionalization and environmental differences, we can generate gradient change data of environmental differences in beekeeping health status. Furthermore, we can analyze the environmental impact characteristics of beekeeping health status based on the gradient change data of environmental differences in beekeeping health status, and generate environmental impact characteristic data of beekeeping health status.

[0046] Furthermore, step S5 includes the following steps:

[0047] Step S51: Analyze the adjustable beekeeping parameters based on the interventionable health status characteristic data, and generate adjustable beekeeping parameter data;

[0048] Step S52: Optimize the spatial design of beekeeping health status by using beekeeping health status-environment related characteristic data to generate beekeeping health status optimization spatial data, and perform global search optimization iteration processing on beekeeping health status optimization spatial data to generate optimized beekeeping health status data.

[0049] Step S53: Design an intelligent management strategy for optimized beekeeping based on optimized beekeeping health status data and adjustable beekeeping parameter data.

[0050] The beneficial effects of this application are as follows: Based on the analysis of bee activity distribution characteristics using related data (including hive structure and activity range data), this invention designs a multi-source sensor topology configuration, achieving targeted and rational sensor deployment and avoiding monitoring blind spots or data redundancy caused by blind deployment. Subsequently, the configured multi-source sensors accurately collect bee behavior perception data (flight images, vibration signals, and acoustic signatures) and regional environmental data (temperature, humidity, and gas composition), and after consistency verification, adjustment, temporal and spatial calibration, effectively improve the standardization and reliability of the multi-source monitoring perception data. Through multi-dimensional hierarchical characteristic segmentation, the spatiotemporally calibrated multi-source data is structured according to data type, monitoring dimension, and importance, allowing different types and dimensions of perception data to be properly positioned, forming standardized and orderly multi-dimensional perception characteristic data of the bee region. Using the multi-dimensional perception characteristic data of the bee region as input, bee behavior sub-domain modal extraction is performed, decomposing the complex multi-source perception data into multiple sub-domain data such as flight mode, vibration mode, and acoustic signature mode, achieving precise segmentation and focusing of different dimensions of bee behavior characteristics. By analyzing the multimodal clustering features of bee behavior, clustering algorithms are used to categorize behavioral data with similar characteristics, extracting different bee behavioral patterns, such as normal foraging behavior, reproductive behavior, and abnormal agitation, clearly defining the feature boundaries of different behavioral patterns. Considering the temporal dynamic characteristics of bee behavior, a dynamic temporal window for bee behavioral events is designed based on the multimodal temporal feature analysis of bee behavior, enabling precise capture of behavioral changes at different time scales. Combined with spatial feature analysis of bee behavioral modalities, the spatial distribution patterns of behavioral data are explored, and then the dynamic temporal window is used to detect behavioral events in the spatial feature data, accurately identifying the occurrence and development process of individual behavioral events, generating bee behavioral event data. Through the extracted bee behavioral pattern data, event combination difference analysis is performed on the behavioral event data, filtering out heterogeneous event combinations with significant health correlations, generating bee behavioral heterogeneous event combination feature data. This approach deeply captures the intrinsic patterns of bee behavior from multiple dimensions, including temporal, spatial, and modal combinations, accurately identifying subtle abnormal signals that are difficult to discern from single behavioral events (such as abnormal combinations of specific vibrations and vocalizations, subtle deviations in flight trajectories, etc.). These subtle signals are often early manifestations of abnormal bee colony health. Using the characteristic data of heterogeneous combinations of bee behavioral events as the core assessment basis, combined with professional knowledge in the field of beekeeping health, targeted assessment and treatment of beekeeping health status are carried out.Unlike existing technologies where beekeepers rely solely on visual observation and subjective judgment of bee colony health (e.g., judging solely by superficial phenomena such as bee activity or the presence of dead individuals), this method uses deeply mined heterogeneous event combinations closely related to bee colony health as the core of the assessment. By establishing quantitative correlations between behavioral characteristics and health status (e.g., the correlation between the frequency of specific heterogeneous event combinations and the probability of disease and pest occurrence, and the correlation between the magnitude of behavioral characteristic changes and bee colony vitality), a quantitative assessment of bee colony health status is achieved. This assessment process is unaffected by subjective human experience or observational skills, objectively and comprehensively reflecting the true health status of the bee colony. The generated beekeeping health status assessment data (e.g., health status, potential risk types, risk severity, etc.) possesses high reliability and accuracy. Furthermore, multidimensional sensory characteristic data of bee regions (including environmental and behavioral data) are linked and analyzed with the beekeeping health status assessment data. Its core focus is on analyzing the correlation between bee colony health status and environmental indicators (temperature, humidity, gas composition, etc.) under different beekeeping behaviors (such as feeding frequency, hive cleaning cycle, and pest and disease control measures). Through correlation modeling and other methods, it uncovers the changing patterns of bee colony health status under different combinations of environmental indicators and different beekeeping behaviors. The bee colony health status-environment related characteristic data generated through this analysis process clearly reveals the influence weights, thresholds, and interaction patterns of environmental indicators and beekeeping behaviors on bee colony health. This not only provides a solid scientific basis for analyzing the causes of fluctuations in bee colony health status (e.g., clarifying whether a decline in health status is due to excessive temperature and humidity or improper beekeeping behavior), but more importantly, it constructs a closed-loop correlation system of "environment-beekeeping behavior-health status." Based on the bee colony health status-environment related characteristic data, intelligent management strategies for optimized beekeeping are designed. For different bee colonies with specific health status and environmental conditions, optimizable beekeeping parameters (such as feed amount, temperature and humidity control targets, hive cleaning frequency, etc.) are precisely identified. By transforming correlational patterns into actionable intelligent management rules, personalized management strategies tailored to the current bee colony are generated. For example, to address health risks caused by excessive temperature and humidity, automatic temperature and humidity control strategies are generated; to address decreased vitality due to improper feeding, precise feeding plans are formulated. This intelligent management strategy possesses dynamic adaptability, capable of dynamically adjusting based on updated monitoring data and evaluation results as environmental conditions change and the health status of the bee colony evolves, achieving precise intervention and dynamic control of the beekeeping process. Furthermore, the generation and execution of intelligent strategies significantly reduce reliance on human experience, minimizing subjectivity and errors in manual management, and improving the efficiency and scientific rigor of beekeeping management.

[0051] Therefore, the AI-based beekeeping health status analysis and optimization management method of this invention addresses the fragmentation and lack of multi-source data integration in existing monitoring methods. It achieves systematic collection of bee behavior perception data and beekeeping environment data through multi-source sensor topology configuration. Combined with data consistency verification, spatiotemporal calibration, and multi-dimensional hierarchical characteristic division, it realizes efficient integration and standardized processing of multi-source perception data, laying a solid foundation for the collaborative analysis of bee behavior characteristics and environmental parameters. Addressing the shortcomings of simple bee behavior analysis and difficulty in identifying early health anomalies, this method deeply explores the combination patterns and evolutionary trends of bee behavior events through bee behavior subdomain modality extraction, multimodal clustering, and heterogeneous event combination feature analysis. This accurately captures early subtle signals reflecting the health status of bee colonies, significantly improving the timeliness and accuracy of health anomaly identification. Addressing the lack of quantitative correlation models between health status and environment / beekeeping behavior, and the inability to adapt to personalized needs, this method constructs a beekeeping health status-environment related feature data system through correlation feature analysis of beekeeping health status and environmental indicators. This provides a scientific quantitative basis for health status assessment, and based on this data system, it can accurately match the beekeeping needs of different bee colonies, achieving personalized health management. To address the lack of intelligent analysis and dynamic optimization mechanisms, this method leverages artificial intelligence technology to achieve intelligent analysis of multi-source data. By combining health status optimization spatial design with global search optimization iteration, it can automatically generate optimized intelligent management strategies for beekeeping, enabling early prediction of health risks and dynamic adaptation of beekeeping management, thus significantly improving the level of intelligence and management efficiency in beekeeping. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the steps of an artificial intelligence-based method for analyzing and optimizing the health status of beekeeping according to the present invention.

[0053] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0054] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0055] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0056] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0057] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for analyzing and optimizing the health status of beekeeping based on artificial intelligence. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of an artificial intelligence-based method for analyzing and optimizing the health status of beekeeping, which includes the following steps:

[0058] Step S1: Use multi-source sensors to perform multi-source monitoring and sensing processing on the target bee area to generate multi-source monitoring and sensing data of the target bee area; use the multi-source monitoring and sensing data of the target bee area to perform multi-dimensional hierarchical characteristic classification processing on the target bee area sensing and monitoring to generate multi-dimensional sensing characteristic data of the bee area.

[0059] In this embodiment of the invention, a comprehensive monitoring network is constructed using multi-source sensors to achieve comprehensive data collection and structured segmentation of bee area data. The deployment of multi-source sensors is planned, and based on the core areas and behavioral characteristics of bee activity, sensors are categorized by function and deployed in the core activity area, transition area at the hive entrance, core foraging area, and peripheral exploration area, ensuring coverage of the entire activity chain of bees from resting and entering / exiting to foraging. Various types of sensors work together to monitor and perceive: image sensors capture image data of bee flight and crawling; vibration sensors collect vibration signals generated by bees making honey and crawling within the hive; acoustic sensors record acoustic signals generated by bee vocalizations and activities; temperature and humidity sensors monitor environmental temperature and humidity data in each area; and gas sensors collect data on key gas components such as carbon dioxide, oxygen, and ammonia. All sensors collect data synchronously in a unified time sequence to ensure spatiotemporal consistency. After collection, the raw monitoring data undergoes basic verification and adjustment, invalid and abnormal data are removed, and missing data is supplemented to ensure that the data accuracy meets the requirements of subsequent analysis. Subsequently, multi-dimensional hierarchical characteristic segmentation was carried out, including data type dimension, spatial hierarchy dimension, and temporal hierarchy dimension. The data type dimension was divided into a bee behavior perception data layer and a bee regional environment data layer. The spatial hierarchy dimension corresponds to the four major areas where the sensors are deployed. The temporal hierarchy dimension is divided into resting periods and active periods according to the bee activity rhythm. By classifying each monitoring data into the corresponding dimension hierarchy according to its attributes, a correlation mapping between data and time, space, and type is established. Finally, multi-dimensional perception characteristic data of bee regions with a clear hierarchical structure is generated, providing a structured data foundation for subsequent behavioral feature analysis.

[0060] Step S2: Perform heterogeneous event combination feature analysis on the multidimensional perception characteristic data of bee regions to generate heterogeneous event combination feature data of bee behavior;

[0061] In this embodiment of the invention, based on the structure of multi-dimensional hierarchical characteristic data, different behavioral subdomains are decomposed according to bee behavior types. Modal extraction is performed on the hierarchical data corresponding to each subdomain. Morphological and motion features of bee behavior are extracted from image data, time-domain and frequency-domain features are extracted from vibration and acoustic signals, and behavior-related correlation features are extracted from environmental data, forming modal feature data for each behavioral subdomain. Subsequently, multimodal clustering processing is performed on the modal feature data of each subdomain. First, feature normalization is used to eliminate the dimensional differences of features in different dimensions, and then a clustering algorithm is used to group modal data with similar features into one category, obtaining different basic bee behavior patterns. Based on the behavior patterns obtained by clustering, behavioral event recognition is carried out in combination with the feature change patterns in the time dimension. By setting a dynamic time window to monitor feature changes, when the feature change meets the preset event judgment criteria, the behavior in that time period is defined as a discrete behavioral event, and the spatiotemporal attributes and feature change information of the event are recorded. Then, the differences in event combinations under normal and abnormal behavior patterns are compared, and the differences in the time sequence patterns, type proportions, and other characteristics of event sequences under different patterns are analyzed. For significantly different event combinations, event features from multimodal sources are integrated to construct heterogeneous event combination features that can characterize the essential attributes of behavior. This results in heterogeneous event combination feature data of bee behavior that includes event combination type, fusion features, and spatiotemporal correlation information, providing core feature support for health status assessment.

[0062] Step S3: Perform beekeeping health status assessment processing on the combined feature data of heterogeneous bee behavior events to generate beekeeping health status assessment data;

[0063] In this embodiment of the invention, a complete health status assessment link is constructed using heterogeneous event combination feature data of bee behavior as the core, enabling accurate determination of the health status of beekeeping. Health status-related feature screening is conducted. Based on the core influencing dimensions of beekeeping health status, the correlation between heterogeneous event combination features and health status is established. Quantitative analysis is used to screen out core correlation features that significantly influence health status, eliminating redundant features to ensure the accuracy of the assessment input. Subsequently, basic activity features representing the basic survival ability of bees and operational behavior features representing production capacity are extracted from the core correlation features. These two types of features are analyzed in detail to clarify the correspondence between each feature indicator and health status, establishing a matching standard between feature indicators and health benchmarks. Based on this, behavioral evolution trend analysis is conducted. The changing patterns of basic activity features and operational behavior features are tracked over time to predict the development trend of bee behavior and identify abnormal signals in trend changes. Simultaneously, historical bee health status assessment data matching the target beekeeping area environment is acquired. Based on this historical data, a multi-level health status discrimination standard system is constructed, clarifying the feature threshold ranges corresponding to different health statuses. The system matches the current core related features and behavioral evolution trend data with the discrimination standard system, calculates the matching degree by weighting, and combines it with abnormal signal verification to determine the health status level of the target bee breeding area, generating bee breeding health status assessment data that includes health status, feature matching details, and trend prediction results.

[0064] Step S4: Based on the multidimensional perception characteristic data of bee areas and the health status assessment data of bee farming, conduct relevant feature analysis on the health status and environmental indicators under different farming behaviors, and generate bee farming health status-environment related feature data.

[0065] This invention focuses on the correlation analysis between multidimensional sensory characteristic data and health status assessment data of bee regions to uncover the inherent correlation between health status and environmental indicators under different beekeeping behaviors. Environmentally relevant data is extracted from the multidimensional sensory characteristic data of bee regions, and environmental distribution characteristic analysis is conducted to identify the distribution patterns of environmental indicators in different spatial regions and time periods, clarifying the spatiotemporal variation characteristics of environmental indicators. Subsequently, based on spatiotemporal synchronization, a correlation mapping is established between beekeeping health status assessment data and environmental distribution characteristic data to ensure accurate matching of health status data and environmental data within the same spatiotemporal unit. The correlation strength between health status and various environmental indicators is calculated through quantitative analysis, and key environmental indicators that significantly affect health status are selected. Based on this, different beekeeping behavior variables are set to construct multiple control groups. Each control group has other conditions fixed, only changing a single beekeeping behavior parameter. The differences in the correlation between key environmental indicators and health status under different beekeeping behaviors are compared and analyzed. By statistically analyzing the distribution range of environmental indicators corresponding to each health status under different beekeeping behaviors, the optimal combination of beekeeping behaviors and environmental indicators that can maintain the health status at the optimal level is identified. Finally, by integrating the analysis results of all control groups, we extracted the correlation characteristics between health status and environmental indicators under different breeding behaviors, identified the environmental indicator range corresponding to the optimal breeding behavior, and generated bee breeding health status-environment related characteristic data that includes breeding behavior type, environmental indicators, correlation strength of health status, and optimal matching combination.

[0066] Step S5: Design intelligent management strategies for optimized beekeeping by using beekeeping health status and environmental-related characteristic data.

[0067] In this embodiment of the invention, based on beekeeping health status-environment related characteristic data, a practical intelligent management strategy for optimized beekeeping is formed. By combining the optimal matching combination from the health status-environment related characteristic data, interventionist features that can be changed through human intervention and have a significant impact on health status are selected. Quantifiable adjustable parameters for beekeeping are extracted from these interventionist features, clarifying the control boundaries, control precision, and interrelationships of each adjustable parameter, thus forming an adjustable parameter system. Subsequently, based on the optimal matching combination, the optimization range of the adjustable parameters is defined, constructing a multi-dimensional optimization space with adjustable parameters as the dimension, clarifying the directions for improvement and optimization goals of each parameter under the current health status. Combining the optimization space and the adjustable parameter system, a modular intelligent management strategy framework is designed, including an environmental control module, a beekeeping behavior control module, and a monitoring and feedback module. The environmental control module clearly defines the triggering conditions, control methods, and control targets for each key environmental indicator, ensuring that these indicators are maintained within their optimal range. The beekeeping behavior control module establishes operational guidelines that match the optimal environmental range for beekeeping behaviors such as comb management, nectar source configuration, and hive layout. The monitoring and feedback module specifies the monitoring frequency, data transmission path, and anomaly response mechanism for each parameter, enabling real-time monitoring of the control effects. Finally, the content of each module is integrated to form a closed-loop intelligent management strategy for optimized beekeeping, ensuring that the strategy can be dynamically adjusted based on environmental changes and health status feedback, achieving continuous optimization of the health status of beekeeping.

[0068] Furthermore, the multi-source monitoring and sensing data of the target bee area mentioned in step S1 includes bee behavior sensing data and bee area environmental data. The bee behavior sensing data includes bee flight image data, bee monitoring vibration signals, and bee monitoring voiceprint signals. The bee area environmental data includes area environmental temperature and humidity data and area environmental gas composition data.

[0069] Furthermore, step S1 includes the following steps:

[0070] Step S11: Design multi-source sensor topology configuration data;

[0071] In this embodiment of the invention, a high-definition camera continuously records the activity trajectories of bees inside and outside the beehive, while a laser rangefinder measures the flight distance and activity range of bees after they enter and leave the beehive, obtaining bee activity correlation data. Based on the above bee activity correlation data, distribution characteristic analysis is performed, and multi-source sensor topology configuration data is designed according to these distribution characteristics.

[0072] Step S12: Use the multi-source sensor topology configuration data to perform multi-source sensor topology configuration operation on the multi-source sensor, and use the configured multi-source sensor to perform multi-source bee monitoring and sensing processing on the target bee area to generate multi-source monitoring and sensing data of the target bee area.

[0073] In this embodiment of the invention, sensor topology configuration is carried out based on multi-source sensor topology configuration data. Specifically, piezoelectric vibration sensors, small acoustic signature sensors, and miniature temperature and humidity sensors are configured at three monitoring points inside the hive; a high-definition infrared image sensor, an acoustic signature sensor, and a temperature and humidity sensor are configured at the monitoring point at the hive entrance; and high-definition infrared image sensors, temperature and humidity sensors, and infrared gas sensors are configured at each monitoring point in the core foraging area and the outer exploration area. During configuration, the sensors are securely installed at the preset monitoring points using bolts, ensuring that the sensors inside the hive do not contact the honeycomb and do not affect bee activity. External sensors are fixed with waterproof brackets to avoid interference from wind and rain. After configuration, all sensors are activated to conduct continuous 72-hour multi-source bee monitoring and sensing operations: the vibration sensor inside the hive collects vibration signals generated by bee activity, and the acoustic signature sensor collects bee vocalization signals; the high-definition infrared image sensor at the hive entrance and external areas collects bee flight images; all temperature and humidity sensors collect data at a frequency of 5 minutes / time, with the measurement range set to temperature 0-50℃ and humidity 30%-95%; and the infrared gas sensor collects carbon dioxide, oxygen, and ammonia concentration data within the area at a frequency of 10 minutes / time. Through the above monitoring operations, multi-source monitoring and sensing data of the target bee area is generated, including bee flight image data, bee monitoring vibration signals, bee monitoring acoustic signals, regional environmental temperature and humidity data, and regional environmental gas composition data.

[0074] Step S13: Perform data consistency verification and adjustment processing on the multi-source monitoring and sensing data of the target bee area to generate standard multi-source monitoring and sensing data of the target bee area;

[0075] In this embodiment of the invention, data consistency verification and adjustment processing is carried out on multi-source monitoring and sensing data of the target bee area to construct a data verification benchmark system. The vibration signal verification benchmark is amplitude 0.1-5V and frequency 10-1000Hz; the acoustic signal verification benchmark is amplitude 0.01-2V and frequency 100-5000Hz; the temperature and humidity data verification benchmark is temperature 15-35℃ and humidity 60-85%; and the gas composition data verification benchmark is carbon dioxide concentration 0.03%-0.1%, oxygen concentration 20%-21%, and ammonia concentration ≤0.001%. Each monitoring data point was compared with its corresponding benchmark range, and abnormal data exceeding the benchmark range was eliminated. For spike pulse interference in the vibration signal, a moving average filtering method was used, with a filtering window length set to 5 sampling points. For environmental noise interference in the acoustic signature signal, wavelet threshold denoising was used, with a wavelet basis of db4 and a decomposition level of 3. For abrupt changes in temperature, humidity, and gas composition data, linear interpolation was used to supplement missing data, with an interpolation interval not exceeding 2 acquisition cycles. Simultaneously, data of the same type collected by different sensors in the same monitoring area were compared, and data deviation values ​​were calculated. When the deviation value exceeded ±2%, the arithmetic mean of data collected by multiple sensors in the area was used to replace the excessively large deviation data. Through the above verification and adjustment processes, standard multi-source monitoring and sensing data for the target bee area was generated, meeting the following accuracy requirements: vibration signal signal-to-noise ratio ≥30dB, acoustic signature signal-to-noise ratio ≥25dB, temperature and humidity measurement error ≤±0.5℃ / ±2%RH, and gas composition measurement error ≤±0.005%.

[0076] Step S14: Perform time-series and spatial calibration processing on the multi-source monitoring and sensing data of the standard target bee area to generate spatiotemporal target bee area multi-source monitoring and sensing data;

[0077] In this embodiment of the invention, time-series and spatial calibration processing is carried out on the multi-source monitoring and sensing data of the standard target bee area. During the time-series calibration process, the Network Time Protocol (NTP) is used to synchronize the acquisition time of all sensors to ensure that the timestamps of the data collected by different sensors at the same moment are completely consistent, and the time synchronization error is controlled within ≤1ms. During the spatial calibration process, based on the sensor topology configuration position determined in step S11, a unique spatial coordinate code is assigned to each sensor. The coordinate code adopts the format of "region number-monitoring point number-sensor type number", where the region number is divided into beehive (01), beehive entrance (02), core foraging area (03), and peripheral exploration area (04). The monitoring point number is arranged in the preset order of each region. The sensor type number is divided into vibration sensor (001), acoustic sensor (002), image sensor (003), temperature and humidity sensor (004), and gas sensor (005). Simultaneously, the bee flight image data collected by the image sensor is bound to the corresponding spatial coordinate code. Through the mapping relationship between image pixel coordinates and actual spatial coordinates, the actual spatial location of the bee in the image is realized, with the mapping ratio set to 1 pixel corresponding to 0.1 centimeters. All standard monitoring data are associated and integrated with the corresponding timestamps and spatial coordinate codes to generate spatiotemporal target bee area multi-source monitoring and sensing data containing time-series information, spatial information, and monitoring data, ensuring that the collection time and location of each data point can be accurately traced during subsequent analysis.

[0078] Step S15: Perform multi-dimensional hierarchical characteristic segmentation processing on the spatiotemporal target bee region multi-source monitoring and sensing data to generate bee region multi-dimensional sensing characteristic data.

[0079] In this embodiment of the invention, based on the temporal and spatial attributes of multi-source monitoring and sensing data of the target bee region in spatiotemporal space, a multi-dimensional hierarchical characteristic classification process is carried out for the target bee region sensing and monitoring. The classification dimensions include data type dimension, spatial hierarchy dimension and temporal hierarchy dimension, and each dimension is further subdivided into characteristic levels. The data types are divided into two dimensions: a bee behavior perception data layer and a bee regional environment data layer. The bee behavior perception data layer includes a flight image behavior data sublayer, a vibration behavior signal sublayer, and a voiceprint behavior signal sublayer. The bee regional environment data layer includes a temperature and humidity environment data sublayer and a gas composition environment data sublayer. The spatial hierarchy is divided into a core activity layer inside the hive (corresponding to monitoring point data inside the hive), a transition layer at the hive entrance (corresponding to monitoring point data at the hive entrance), a core foraging activity layer (corresponding to monitoring point data in the core foraging area), and a peripheral exploration activity layer (corresponding to monitoring point data in the peripheral exploration area). The temporal hierarchy is divided into a resting period layer at dawn (00:00-06:00), a morning activity period layer (06:00-12:00), an afternoon activity period layer (12:00-18:00), and a resting period layer at night (18:00-24:00). During the segmentation process, based on the timestamp, spatial coordinate encoding, and data type of each spatiotemporal monitoring data point, it is categorized into the corresponding level of the corresponding dimension. For example, vibration signal data collected from monitoring points inside beehives during the early morning hours is categorized into the vibration behavior signal sub-layer of the data type dimension, the core activity layer inside the beehive of the spatial level dimension, and the early morning resting period layer of the temporal level dimension. The data at each level is then labeled and integrated to generate multidimensional perception characteristic data of bee regions containing multidimensional hierarchical labeling information and corresponding monitoring data. This results in a clear spatiotemporal-type-hierarchical association structure, providing a structured data foundation for subsequent analysis of heterogeneous event combinations in bee behavior.

[0080] Furthermore, step S11 includes the following steps:

[0081] Analyze bee activity correlation data in the target bee area to generate bee activity correlation data, wherein the bee activity correlation data includes beehive structure data and bee activity range data.

[0082] Based on the correlation data of bee activity, the distribution characteristics of bee activity are analyzed to generate bee activity distribution characteristic data, and multi-source sensor topology configuration data is designed based on the bee activity distribution characteristic data.

[0083] In this embodiment of the invention, bee activity correlation data analysis is conducted on the target bee area. A full-area monitoring network is constructed by deploying high-definition infrared cameras and laser rangefinders. Four high-definition infrared cameras are evenly deployed 1 meter around the beehive, each with a frame rate of 30 frames / second and a resolution of 1920×1080 pixels. The laser rangefinders are deployed 2 meters directly in front of the beehive entrance and exit, with a sampling frequency of 10 times / second. The monitoring period lasts for 30 days, achieving continuous recording of bee activities inside and outside the beehive throughout the entire time. During the monitoring process, the cameras capture the trajectories, stopping positions, and colony aggregation states of bees entering and leaving the beehive. Simultaneously, the laser rangefinders measure the bee flight distance and activity radius. Combined with the actual structure of the beehive, the distribution of bee activity areas within the beehive is recorded. Bee activity correlation data is generated based on the monitoring data. Based on the generated bee activity correlation data, bee activity distribution characteristics are analyzed. A distribution characteristic evaluation index system is constructed by statistically analyzing the number of bees, activity frequency, and dwell time in different areas per unit time. The average and fluctuation range of bee activity density in each area are calculated using statistical data from 30 consecutive days. Based on this distribution characteristic data, a multi-source sensor topology configuration was designed. Areas with higher activity density had higher sensor deployment density and more comprehensive monitoring dimensions to ensure accurate and comprehensive data collection. For example, three sensor monitoring points were evenly distributed 10 cm above the honeycomb inside the hive, corresponding to the left, middle, and right sides of the honeycomb, with each monitoring point covering two honeycomb areas. A comprehensive monitoring point was placed 50 cm outside the hive entrance, covering the entire entrance and exit area and a 1-meter radius around it. In the core foraging area, a sensor array was deployed at 5m x 5m intervals, with a total of 20 monitoring points, ensuring each monitoring point covered an area of ​​25 square meters without blind spots. In the outer exploration area, sensors were deployed at 10m x 10m intervals, with a total of 15 monitoring points, covering the entire outer area. Each monitoring point has a clearly defined combination of sensor types. The monitoring points inside the beehive are equipped with piezoelectric vibration sensors, small acoustic fingerprint sensors, and miniature temperature and humidity sensors. The external monitoring points are equipped with high-definition infrared image sensors, temperature and humidity sensors, and infrared gas sensors according to the area's function. The sensors at the monitoring points inside the beehive are installed 30 centimeters above the bottom of the beehive, and the sensors at the external monitoring points are installed 1.5 meters above the ground to ensure that the sensor's collection range completely overlaps with the bee's activity area.

[0084] Furthermore, step S2 includes the following steps:

[0085] Step S21: Extract bee behavior subdomain modalities based on bee region multidimensional perception characteristic data to generate bee behavior subdomain modal data;

[0086] In this embodiment of the invention, based on the multidimensional hierarchical structure of bee regional multidimensional perception characteristic data, bee behavior subdomain modal extraction is carried out. The extraction logic is to divide the behavior subdomains according to the data type dimension and behavior association characteristics. Each subdomain corresponds to a specific bee behavior type, and exclusive modal features are extracted from the corresponding hierarchical data. The divided behavior subdomains include flight behavior subdomain, crawling behavior subdomain, honey-collecting behavior subdomain, honey-making behavior subdomain, and resting behavior subdomain. Among them, the flight behavior subdomain corresponds to the morning and afternoon activity time data of the flight image behavior data sublayer, the crawling behavior subdomain corresponds to the core activity layer data inside the beehive of the flight image behavior data sublayer, the honey-collecting behavior subdomain corresponds to the core foraging area data of the flight image behavior data sublayer, the honey-making behavior subdomain corresponds to the core activity layer data inside the beehive of the vibration behavior signal sublayer, and the resting behavior subdomain corresponds to the early morning and night resting time data of the voiceprint behavior signal sublayer. During the extraction process, the flight behavior subdomain modality extraction uses a convolutional neural network to extract the contour features, motion trajectory features, and velocity features of bees in the image. The convolutional kernel size is set to 3×3, the stride is 1, and the extracted feature dimension is 128. The crawling behavior subdomain modality extraction uses morphological processing to extract the crawling path and body posture features of bees in the image. The morphological erosion kernel size is 2×2, the dilation kernel size is 2×2, and the feature dimension is 64. The honey-gathering behavior subdomain modality extraction uses a fusion extraction of color features and texture features. The color features use 36-dimensional features in HSV space, and the texture features use 16-dimensional features in gray-level co-occurrence matrix, resulting in a fusion feature dimension of 52. The honey-making behavior subdomain modality extraction uses a fusion of time-domain and frequency-domain features of vibration signals. The time-domain features include peak value, mean, and variance, and the frequency-domain features include dominant frequency and frequency band energy, resulting in an 8-dimensional feature dimension. The resting behavior subdomain modality extraction uses the spectral features of the voiceprint signal, with the extracted frequency band being 100-1000Hz, resulting in a 32-dimensional feature dimension. The modal features extracted from each subdomain are associated with the corresponding spatial and temporal hierarchical information to generate bee behavior subdomain modal data containing subdomain type identifiers, modal feature vectors, and spatiotemporal information.

[0087] Step S22: Perform multimodal clustering feature analysis on the bee behavior subdomain modal data to generate multimodal clustering feature data on bee behavior, and perform bee behavior pattern analysis based on the multimodal clustering feature data on bee behavior to generate bee behavior pattern data;

[0088] In this embodiment of the invention, multimodal clustering feature analysis of bee behavior subdomain modal data is performed. Before clustering, the modal feature vectors of each subdomain are normalized, with the normalization range set to [0,1]. The min-max normalization method is used to eliminate the dimensional differences of features in different dimensions. The K-means clustering algorithm is used, with 5 clusters corresponding to the 5 basic bee behavior types. The initial cluster centers are determined by randomly selecting 5 sets of modal feature vectors from different subdomains. The iteration termination condition is set to 100 iterations or a change in cluster centers of less than 0.001. During the clustering process, the Euclidean distance between each modal feature vector and each cluster center is calculated, and the feature vector is assigned to the nearest cluster. After each iteration, the cluster center is updated to the mean of all feature vectors in that category. The clustering analysis is completed through multiple iterations, generating multimodal clustering feature data of bee behavior. This data contains 5 cluster categories, each category corresponding to a set of cluster center feature vectors and all modal feature vectors under that category. Based on clustering results, bee behavior pattern analysis was conducted. Behavior pattern types were determined by comparing the matching degree between the feature vectors of each cluster center and known bee behavior characteristics. The first cluster category corresponds to the normal flight mode, with a movement speed feature value of 0.5-0.8 m / s and a contour integrity feature value ≥0.9 in the cluster center feature vector. The second cluster category corresponds to the abnormal flight mode, with a movement speed feature value <0.2 m / s or >1.0 m / s and a contour integrity feature value <0.6. The third cluster category corresponds to the normal honey-making mode, with a vibration signal dominant frequency feature value of 50-80 Hz and frequency band energy concentrated in 20-100 Hz. The fourth cluster category corresponds to the normal resting mode, with the acoustic signature signal spectrum energy concentrated in 100-300 Hz and amplitude fluctuation ≤0.1V. The fifth cluster category corresponds to the abnormal resting mode, with the acoustic signature signal spectrum energy dispersed and amplitude fluctuation >0.3V. Based on this, bee behavior pattern data containing behavior pattern type, corresponding cluster characteristics, and spatiotemporal distribution information was generated.

[0089] Step S23: Perform bee behavior event analysis on the bee behavior subdomain modal data to generate bee behavior event data;

[0090] In this embodiment of the invention, bee behavior event analysis is performed on bee behavior subdomain modal data. The analysis logic is to identify discrete behavior events based on the temporal variation patterns of modal features. A behavior event is defined as a behavioral unit where modal features show significant changes and the duration reaches a set threshold. Based on the temporal hierarchy attributes of the bee behavior subdomain modal data, a dynamic temporal window for bee behavior events is designed, with a window length of 10 seconds and a window overlap rate of 50% to ensure no behavioral changes are missed. For flight behavior subdomain modal data, changes in motion velocity and contour features within the window are monitored. When the velocity feature change is >0.3 m / s and lasts for two consecutive windows, it is determined to be a flight state change event. For honey-making behavior subdomain modal data, changes in the dominant frequency characteristics of vibration signals within the window are monitored. When the dominant frequency characteristic change is >20 Hz and lasts for three consecutive windows, it is determined to be a honey-making intensity change event. For resting behavior subdomain modal data, changes in the amplitude fluctuation characteristics of voiceprint signals within the window are monitored. When the fluctuation characteristic change is >0.2V and lasts for one window, it is determined to be a resting state disturbance event. Simultaneously, by combining spatial hierarchy information, events are bound to corresponding spatial regions. For example, changes in vibration signals in the core activity layer within the beehive are identified as honey-making events within the beehive, and changes in flight speed in the core foraging area are identified as foraging flight events. All identified events are labeled, and the start time, end time, occurrence area, and characteristic change amplitude are recorded to generate bee behavior event data containing event identifiers, event types, spatiotemporal parameters, and characteristic change data.

[0091] Step S24: Perform event combination difference analysis on bee behavior event data using bee behavior pattern data to generate bee behavior event combination difference data;

[0092] In this embodiment of the invention, a difference analysis of event combinations is conducted on bee behavioral event data using bee behavior pattern data. The core of the analysis is to compare the differences in the temporal sequence, event type proportion, and characteristic change patterns of event combinations under different behavioral patterns. First, bee behavior pattern data is divided into a normal behavior pattern set and an abnormal behavior pattern set. The normal behavior pattern set includes normal flight, normal honey production, and normal resting patterns, while the abnormal behavior pattern set includes abnormal flight and abnormal resting patterns. Bee behavioral event data corresponding to the two pattern sets are extracted to construct event combination sequences. The sequences are arranged chronologically, with each sequence consisting of 20 consecutive events. Difference indices between the normal and abnormal event combination sequences are calculated, including event type overlap, variance of time intervals between adjacent events, and mean characteristic change amplitude. The event type overlap is calculated using the Jaccard coefficient, the time interval variance is calculated based on the start time difference between adjacent events, and the mean characteristic change amplitude is the arithmetic mean of the characteristic changes of all events within the sequence. A threshold for determining differences was set: event type overlap < 0.6, time interval variance > 5 seconds², and mean characteristic variation amplitude > 0.3. Meeting any one of these conditions was considered a significant difference. Analysis showed that normal event combination sequences exhibited a regular time sequence of "normal flight - honey collection - normal flight - normal honey production - normal rest," with a time interval variance of 2-3 seconds² and a mean characteristic variation amplitude of 0.1-0.2. Abnormal event combination sequences exhibited a disordered time sequence of "abnormal flight - resting state disturbance - abnormal flight," with a time interval variance of 6-8 seconds² and a mean characteristic variation amplitude of 0.4-0.5. Based on this, bee behavior event combination difference data was generated, including difference types, difference index values, and comparative data on event combination sequences.

[0093] Step S25: Perform heterogeneous event combination feature analysis on bee behavior based on the data of differences in bee behavior event combinations, and generate feature data of heterogeneous event combinations of bee behavior.

[0094] In this embodiment of the invention, heterogeneous event combination feature analysis of bee behavior is carried out based on the differential data of bee behavior event combinations. A heterogeneous event combination is defined as a combination of event types corresponding to different behavioral subdomain modalities. The analysis logic is to integrate the differential features of multimodal events and construct a comprehensive feature vector that can characterize the health status of bee behavior. First, heterogeneous event combinations corresponding to significant differences in bee behavior event combination differential data are screened, and the event types and modal sources included in each heterogeneous combination are clarified. For example, "abnormal flight event (image modality) - resting state disturbance event (voiceprint modality) - vibration abnormal event (vibration modality)" constitutes a typical abnormal heterogeneous event combination, and "normal flight event (image modality) - honey collection event (image modality) - normal honey making event (vibration modality)" constitutes a typical normal heterogeneous event combination. For each heterogeneous event combination, fusion features were extracted. The fusion dimensions included event type sequence features, fusion values ​​of the variation amplitudes of each modality, event duration percentage, and spatial distribution concentration. The event type sequence features were converted into a 64-dimensional vector using one-hot encoding. The fusion values ​​of the variation amplitudes of each modality were calculated using weighted summation, with image modality weighting at 0.4, vibration modality weighting at 0.3, and acoustic signature weighting at 0.3. The event duration percentage was the ratio of the total duration of all events within the combination to the analysis period (1 hour). The spatial distribution concentration was represented by the ratio of the number of event occurrence areas to the total number of monitored areas. These fusion features were integrated into a 128-dimensional feature vector, with each feature vector corresponding to a heterogeneous event combination type. The difference type (normal / abnormal) of the combination was also labeled, generating heterogeneous event combination feature data for bee behavior, including the heterogeneous event combination type, the 128-dimensional fusion feature vector, and the difference label. This provides core feature basis for subsequent assessment of beekeeping health status.

[0095] Furthermore, step S23 includes the following steps:

[0096] Step S231: Perform multimodal temporal feature analysis of bee behavior based on the bee behavior subdomain modal data, generate multimodal temporal feature data of bee behavior, and design a dynamic temporal window for bee behavior events based on the multimodal temporal feature data of bee behavior;

[0097] In this embodiment of the invention, multimodal temporal feature analysis of bee behavior is carried out based on bee behavior subdomain modal data. The analysis logic is to extract the pattern of change of each subdomain modal feature over time and construct a temporal feature sequence to reflect the dynamic evolution process of bee behavior. For image-based modal data corresponding to the flight behavior, crawling behavior, and honey-gathering behavior subdomains, modal features of each frame of image were extracted in time-stamp order to form a time series. The time series length was set to 60 seconds, the sampling interval was 1 second, and each series contained 60 modal feature vectors. The extracted temporal features included the rate of change of the feature mean, the frequency of the feature peak, and the feature fluctuation period. For vibration modal data of the honey-making behavior subdomain, a time series was constructed according to the timestamps corresponding to a sampling rate of 1000Hz, with a sequence length of 60,000 sampling points. The extracted temporal features included the moving average of the temporal features, the variance of the difference between adjacent sampling points, and the drift of the dominant frequency over time. For acoustic modal data of the resting behavior subdomain, a time series was constructed according to a sampling rate of 44.1kHz, with a sequence length of 2,646,000 sampling points. The extracted temporal features included the temporal variation curve of spectral energy, the time interval of amplitude peaks, and the rate of change of the energy proportion of the feature frequency band. The temporal features of each subdomain are integrated to generate multimodal temporal feature data of bee behavior, including subdomain type, temporal feature sequence, and time span information. Based on this data, a dynamic temporal window for bee behavior events is designed. The window length is determined according to the change cycle of the temporal features of different subdomains. The feature change cycle of the flying, crawling, and honey-collecting behavior subdomains is 10 seconds, so the window length is set to 10 seconds; the feature change cycle of the honey-making behavior subdomain is longer, so the window length is set to 15 seconds; and the feature change cycle of the resting behavior subdomain is shorter, so the window length is set to 5 seconds. The window overlap rate of all subdomains is uniformly set to 50% to ensure that adjacent windows can be connected and that no dynamic changes in behavior are missed.

[0098] Step S232: Perform bee behavior modality spatial feature analysis on the bee behavior subdomain modal data to generate bee behavior modality spatial feature data;

[0099] In this embodiment of the invention, spatial feature analysis of bee behavior modal data is performed on the bee behavior subdomain modal data. The core of the analysis is to explore the correlation between the modal features of each subdomain and spatial location, and to extract features that can reflect the spatial distribution pattern of bee behavior. For the image modal data of the flight behavior subdomain, combined with spatial coordinate encoding, the spatial location coordinates of bees in the image, the spatial curvature of the flight trajectory, and the spatial aggregation degree of the flock flight are extracted. The spatial location coordinates are obtained through the mapping relationship between image pixel coordinates and actual spatial coordinates, with a mapping ratio of 1 pixel to 0.1 cm. The spatial aggregation degree is calculated by the proportion of bees in a unit spatial area, with the unit spatial area set to 10 cm × 10 cm. For the image modal data of the crawling behavior subdomain, the spatial distribution area of ​​bees in the beehive, the spatial span of the crawling path, and the spatial overlap of adjacent bee crawling trajectories are extracted. The spatial distribution area is divided into three sub-regions: left, middle, and right, according to the position of the beehive combs. The degree represents the straight-line distance between the starting and ending points of the crawling trajectory. For the image modal data of the nectar-foraging behavior subdomain, the spatial distribution density of bees in the core foraging area, the spatial distance between the nectar-foraging location and the nectar source plant, and the spatial diffusion range of the nectar-foraging colony are extracted. For the vibration modal data of the honey-making behavior subdomain, combined with the spatial coordinates of sensors within the beehive, the spatial propagation attenuation coefficient of the vibration signal and the phase difference of the vibration signal at different monitoring points are extracted. The propagation attenuation coefficient is the ratio of the vibration amplitude at different spatial locations, and the phase difference accuracy is set to 0.01 radians. For the acoustic modal data of the resting behavior subdomain, the spatial propagation direction of the acoustic signal and the amplitude difference of the acoustic signal at different monitoring points are extracted. The propagation direction is calculated using the arrival time difference of the acoustic signals from multiple monitoring points. The spatial features of each subdomain are integrated to generate spatial feature data of bee behavior modalities.

[0100] Step S233: Detect bee behavior events using the dynamic temporal window of bee behavior events to generate bee behavior event data.

[0101] In this embodiment of the invention, a dynamic temporal window for bee behavior events is used to detect bee behavior events based on spatial features of bee behavior modal spatial feature data. The detection logic is to determine whether there are significant changes in spatial features within the window. If the change meets a set threshold, it is determined that a behavior event has occurred. For the flight behavior subdomain, a 10-second dynamic temporal window is used to traverse the spatial feature data, monitoring the changes in spatial curvature and colony aggregation of bee flight trajectories within the window. When the change in spatial curvature is >0.5 radians / meter and the change in colony aggregation is >30%, it is determined to be a flight trajectory abrupt change event. At the same time, the spatial coordinates of the event and the start and end times are recorded. For the crawling behavior subdomain, a 10-second window is used to monitor the changes in spatial span and trajectory overlap of crawling paths. When the change in spatial span is >5 centimeters and the change in trajectory overlap is >40%, it is determined to be a crawling area transfer event. For the nectar foraging behavior subdomain, a 10-second window is used to monitor the changes in spatial distribution density and diffusion range of nectar foraging. For the honeybee behavior subdomain, changes in distribution density exceeding 2 bees / m² and changes in diffusion range exceeding 1 m² were identified as honeybee migration events. For the honey-making behavior subdomain, changes in the spatial propagation attenuation coefficient and phase difference of vibration signals were monitored within a 15-second window. Changes in the attenuation coefficient exceeding 0.2 and changes in the phase difference exceeding 0.1 radians were identified as spatial differences in honey-making intensity. For the resting behavior subdomain, changes in the spatial propagation direction and amplitude difference of voiceprint signals were monitored within a 5-second window. Changes in the propagation direction exceeding 30 degrees and changes in the amplitude difference exceeding 0.2V were identified as resting spatial disturbance events. These bee behavior event data provide fundamental data support for subsequent analysis of event combination differences.

[0102] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S3 is provided in this embodiment. Step S3 includes:

[0103] Step S31: Perform a correlation feature analysis on the health status of beekeeping based on the combined feature data of heterogeneous events in bee behavior, and generate correlation feature data on the health status of beekeeping.

[0104] In this embodiment of the invention, a correlation feature analysis of bee health status is conducted based on heterogeneous event combination feature data of bee behavior. The core logic is to screen core features that can directly characterize bee health status and establish an intrinsic correlation between heterogeneous event combination features and bee health status. First, the core influencing dimensions of bee health status are identified, including three dimensions: bee colony activity, behavioral coordination, and energy metabolism efficiency. Based on these, a correlation feature screening system is constructed. For the 128-dimensional heterogeneous event combination feature vector generated in S25, the Pearson correlation coefficient is used to calculate the correlation between each feature dimension and the preset health status benchmark indicators. The health status benchmark indicators are set as follows: bee daily flight time ≥ 8 hours, honey-making vibration signal main frequency stable at 50-80Hz, and resting period voiceprint amplitude fluctuation ≤ 0.1V. During the correlation calculation, the analysis period was set to 24 hours. For each feature dimension, three consecutive periods of correlation coefficients were calculated, and the average was taken as the final correlation coefficient value. A correlation coefficient threshold of 0.7 was set, resulting in 42 feature dimensions with correlation coefficients ≥ 0.7. These included 16 dimensions of event type sequence features, 8 dimensions of fusion values ​​of modal feature variation amplitudes, 10 dimensions of event duration proportions, and 8 dimensions of spatial distribution concentration. Redundancy was eliminated from the 42 selected features. The variance inflation factor (VIF) was used to detect multicollinearity among features, with a VIF threshold of 5. Six redundant features with VIF > 5 were removed, ultimately retaining 36 core features. These 36 core features were then integrated with corresponding heterogeneous event combination types and difference identifiers to generate beekeeping health status correlation feature data containing feature names, feature values, correlation coefficients, and their respective health impact dimensions. This provides accurate input for subsequent basic activity and operational behavior feature analysis.

[0105] Step S32: Analyze the basic activity characteristics and operational behavior characteristics of bees based on the correlation characteristics data of bee health status;

[0106] In this embodiment of the invention, basic activity characteristic data and operational behavior characteristic data of bees are analyzed based on the correlation characteristic data of bee health status. The analysis logic is based on the classification of health impact dimensions, and the characteristic indicators related to basic activity and operational behavior are decomposed and refined from the correlation characteristics. For the analysis of basic activity characteristics, the focus is on the basic behavioral capabilities for bee colony survival, and three core indicators are extracted from the correlation characteristics: flight event frequency, resting state stability, and crawling trajectory continuity. Among them, the flight event frequency is calculated as the number of occurrences of heterogeneous flight events per unit time (1 hour), and statistics are collected separately for the morning (06:00-12:00) and afternoon (12:00-18:00) periods; the resting state stability is calculated by the duration fluctuation coefficient of heterogeneous resting events, where the fluctuation coefficient = standard deviation / mean, and the value range is 0-1; the crawling trajectory continuity is calculated by the spatial trajectory overlap rate corresponding to heterogeneous crawling events, where the overlap rate = the overlap length of adjacent trajectories / the total trajectory length. Based on the analysis of operational behavior characteristics, this study focuses on the efficiency of bee production-related behaviors and extracts three core indicators: honey gathering event density, honey production intensity stability, and honey gathering-honey production behavior coordination efficiency. Honey gathering event density is calculated as the number of heterogeneous honey gathering events per hour per unit area (1 square meter) within the core foraging area; honey production intensity stability is calculated by the fusion value of the vibration characteristic change amplitude corresponding to heterogeneous honey production events, with a stability threshold of 0.2, and the percentage of durations with a fusion value ≤ 0.2 is statistically analyzed; honey gathering-honey production behavior coordination efficiency is calculated as the average time interval between the end of a honey gathering event and the start of a honey production event. All indicators were quantified, with the health benchmarks for flight event frequency being ≥30 times / hour in the morning and ≥25 times / hour in the afternoon; the stability fluctuation coefficient of the resting state being ≤0.3; the continuous overlap rate of crawling trajectory being ≥0.6; the honey gathering event density being ≥5 times / square meter·hour; the stability ratio of honey production intensity being ≥80%; and the honey gathering-honey production connection time interval being ≤10 minutes. Based on these, basic bee activity characteristic data containing the quantitative values ​​of each indicator, time period distribution, and health benchmark matching degree were generated, as well as bee operational behavior characteristic data containing operational efficiency indicators, spatial distribution, and temporal connection characteristics.

[0107] Step S33: Analyze the evolutionary trend of bee behavior based on the basic activity characteristics data and operational behavior characteristics data of bees, and generate bee behavior evolutionary trend data;

[0108] In this embodiment of the invention, bee behavior evolution trend analysis is conducted on basic activity characteristic data and bee operational behavior characteristic data. The core logic is to track the changing patterns of these two types of characteristic data based on a time-series dimension to predict the development trend of bee behavior. First, the time span for trend analysis is set to 7 days, with the analysis period divided by days. Each day, the core indicators of basic activity and operational behavior are extracted and quantified to construct 7 sets of time-series characteristic sequences. A sliding window method combined with a linear regression model is used for trend fitting. The sliding window length is set to 3 days, and the window sliding step size is 1 day. Within each window, the least squares method is used to fit the linear regression equation y=kx+b, where x is the time series (1-7 days), y is the characteristic indicator value, and the slope k is used as the trend change rate to determine the trend type. Trend judgment thresholds are set: k>0.05 is judged as an upward trend, -0.05≤k≤0.05 is judged as a stable trend, and k<-0.05 is judged as a downward trend. For basic activity characteristics, the focus is on analyzing the temporal trends of flight event frequency and resting state stability. For example, if the slope k of flight event frequency is less than -0.08 for three consecutive windows, and the slope k of resting state stability fluctuation coefficient is greater than 0.06, then basic activity is considered to be significantly decreasing. For operational behavior characteristics, the focus is on analyzing the temporal trends of honey gathering event density and honey production intensity stability. If the slope k of honey gathering event density is greater than 0.07 for three consecutive windows, and the slope k of honey production intensity stability percentage is greater than 0.05 for three consecutive windows, then operational efficiency is considered to be significantly increasing. Simultaneously, the daily time-level characteristics are considered to analyze the trend differences between morning and afternoon periods. For example, if the flight event frequency trend is stable in the morning and decreasing in the afternoon, it is marked as a periodic activity decline. Finally, all trend judgment results are integrated to generate bee behavior evolution trend data.

[0109] Step S34: Obtain historical bee health status assessment data;

[0110] In this embodiment of the invention, historical bee health status assessment data is obtained. The data source is limited to monitoring and assessment records from the same beekeeping base over the past three years that are completely consistent with the environmental conditions, bee species, and beekeeping scale of the target beekeeping area. Historical data includes health status assessment results, divided into three levels according to a unified standard: healthy level (normal bee activity, stable operational efficiency, no signs of disease), sub-healthy level (slightly decreased activity or fluctuating operational efficiency, potential health risks), and abnormal level (significantly decreased activity, operational stagnation, obvious disease symptoms). The screened historical data is preprocessed to remove extreme data with abnormal fluctuations (using the 3σ criterion to remove data exceeding the mean ± 3 standard deviations). This results in a historical dataset containing multiple sets of valid samples, which is stored in a structure of "heterogeneous event combination characteristics - basic activity characteristics - operational behavior characteristics - evolutionary trend characteristics - health assessment results." Simultaneously, environmental information such as season, temperature, and humidity corresponding to the samples is labeled, providing sufficient and effective data support for the subsequent design of a multi-level discrimination matrix.

[0111] Step S35: Design a multi-level discrimination matrix relationship for bee breeding health status based on historical bee health status assessment data, bee operational behavior data, and bee behavior evolution trend data, and generate a multi-level discrimination matrix for bee breeding health status.

[0112] In this embodiment of the invention, a multi-level discrimination matrix relationship for bee health status is designed based on historical bee health status assessment data, including basic bee activity characteristics, bee operational behavior characteristics, and bee behavior evolution trend data. The core logic is to construct a mapping relationship between feature indicators and health status based on historical samples, forming a three-level discrimination standard system. First, the dimensional structure of the discrimination matrix is ​​determined. The matrix rows are the 12 core feature indicators extracted from S32 and S33 (6 basic activity indicators and 6 operational behavior indicators), and the columns are the three levels of health status (healthy, sub-healthy, and abnormal). The matrix cells represent the threshold range of the corresponding feature indicator under that health status. The threshold range is determined using statistical analysis. The values ​​of each feature indicator in the historical data under different health statuses are statistically analyzed, and the 25th and 75th quantiles for each level are calculated. The threshold range is set to [25th quantile, 75th quantile], ensuring that this range includes 75% of the samples at the corresponding level. The feature threshold range and weight coefficients are integrated into a 12-row, 3-column multi-level discrimination matrix for beekeeping health status. The matrix also includes auxiliary information such as feature indicator names, health status identifiers, and threshold confidence (calculated based on the historical sample matching accuracy, with a confidence level of ≥90%) to ensure the reliability and interpretability of the discrimination matrix.

[0113] Step S36: Based on the multi-level discrimination matrix of bee breeding health status, perform bee breeding health status assessment processing on the combination feature data of heterogeneous events of bee behavior to generate bee breeding health status assessment data.

[0114] In this embodiment of the invention, the health status of beekeeping is assessed based on the multi-level discrimination matrix of beekeeping health status, using heterogeneous event combination feature data of bee behavior. The core logic is to match the current feature data with the thresholds of the discrimination matrix one by one, calculate the matching degree of health status by combining weight correction, and finally determine the assessment result. First, 36-dimensional core correlation features are extracted from the heterogeneous event combination feature data of bee behavior and mapped to 12 core feature indicators corresponding to the discrimination matrix, and the current quantitative value of each indicator is obtained. Each indicator value is compared with the threshold range of the three health statuses in the discrimination matrix, and the matching result is recorded (complete match is recorded as 1, partial match as 0.5, and no match as 0). Combining the weight coefficients corresponding to the evolutionary trend features, the matching results of each health status are weighted and summed to calculate the comprehensive matching score of each level. If there is a situation where the scores of different levels are close (difference < 1), a secondary verification mechanism is activated to supplement the comparison of the difference indicators in the heterogeneous event combination features. If the difference indicator is normal, the level is corrected upward; if the difference indicator is abnormal, the level is corrected downward. During the assessment process, the matching details, weight correction process, and score calculation results of each feature indicator are recorded simultaneously to generate beekeeping health status assessment data. The health assessment level is clearly marked as healthy, sub-healthy, or abnormal. The comprehensive matching score is retained to two decimal places to ensure the accuracy and traceability of the assessment results.

[0115] Furthermore, step S4 includes the following steps:

[0116] Step S41: Analyze the environmental distribution characteristics of bees based on the multidimensional perception characteristic data of bee regions, and generate bee environmental distribution characteristic data;

[0117] In this embodiment of the invention, the environmental distribution characteristics of bees are analyzed based on multidimensional sensing characteristic data of bee regions. Spatial distribution patterns and temporal variation characteristics are extracted from the multidimensional environmental data, and a correlation mapping between environmental indicators and bee activity areas is established. The environmental data in the multidimensional sensing characteristic data of bee regions comes from the temperature and humidity environmental data sub-layer and the gas composition environmental data sub-layer, covering four spatial regions: the core activity layer inside the hive, the transition layer at the hive entrance, the core foraging activity layer, and the outer exploration activity layer, as well as four time periods: early morning, morning, afternoon, and night. During the analysis, analysis units are constructed according to the intersection dimension of spatial regions and time periods, resulting in a total of 4×4=16 analysis units. Each unit collects environmental data for 7 consecutive days. The extracted environmental distribution characteristics include spatial distribution characteristics and temporal variation characteristics. Spatial distribution characteristics are determined by calculating the mean, maximum, minimum values, and spatial coefficient of variation of environmental indicators for each region. The spatial coefficient of variation = standard deviation of indicators at each monitoring point within the region / regional mean. Temporal variation characteristics are determined by calculating the amplitude, rate of change, and fluctuation period of environmental indicators for each time period. The amplitude of change = absolute value of the difference between indicators in adjacent time periods within the same region; the rate of change = amplitude of change / time interval; and the fluctuation period is obtained through statistical analysis of peak intervals in time-series data. Specific environmental indicators include temperature, humidity, carbon dioxide concentration, oxygen concentration, and ammonia concentration. The statistical precision for temperature is 0.1℃, humidity is 0.1%RH, and gas concentration is 0.001%. For example, the average temperature in the core activity layer of the beehive during the morning is 28.5℃ with a spatial coefficient of variation of 0.02, while the temperature variation amplitude during the afternoon is 2.3℃ with a rate of change of 0.38℃ / hour; the average carbon dioxide concentration in the core foraging area is 0.045% with a spatial coefficient of variation of 0.05. The environmental distribution characteristics of all analysis units are associated with corresponding spatial and temporal codes to generate bee environmental distribution characteristic data that includes spatial regions, time periods, environmental indicator types, spatial distribution parameters, and temporal variation parameters.

[0118] Step S42: Perform health status and environmental correlation processing on the beekeeping health status assessment data and bee environmental distribution characteristic data to generate beekeeping health status-environment correlation data;

[0119] In this embodiment of the invention, beekeeping health status assessment data and bee environmental distribution characteristic data are processed to correlate beekeeping health status with the environment. A spatiotemporal correlation mapping between health assessment results and environmental indicators is established to quantify the impact of different environmental conditions on health status. Based on spatiotemporal synchronization, the timestamps and spatial coordinate codes of the beekeeping health status assessment data are matched with the corresponding information of the bee environmental distribution characteristic data to ensure accurate correlation between health data and environmental data in the same spatiotemporal unit, with matching errors controlled within ≤1 minute and ≤0.5 meters. After correlation, a two-dimensional data matrix containing health status and environmental indicators is constructed, with 16 spatiotemporal analysis units in the row dimension and health assessment levels (healthy, sub-healthy, abnormal) and 5 environmental indicators in the column dimension. Based on this matrix, the correlation strength between health status and each environmental indicator is calculated, and Spearman's rank correlation coefficient is used for quantification. The correlation coefficient ranges from [-1, 1], with a larger absolute value indicating a higher correlation strength. A correlation strength threshold of 0.6 was set, and environmental indicators with an absolute correlation strength ≥ 0.6 were selected. For example, the correlation coefficient between hive temperature and health status was -0.78 (both excessively high and low temperatures lead to a decline in health status), the correlation coefficient between humidity in the core foraging area and health status was 0.72 (health status is highest when humidity is between 65% and 75%), and the correlation coefficient between hive ammonia concentration and health status was -0.83 (health status significantly declines when ammonia concentration exceeds 0.001%). The distribution range of each environmental indicator under different health statuses was also statistically analyzed. The correlation strength data, the distribution range of environmental indicators, and the corresponding spatiotemporal units and health status were integrated to generate beekeeping health status-environment correlation data.

[0120] Step S43: Based on the correlation data between beekeeping health status and environment, conduct health status intervention feature analysis on beekeeping behavior and environmental indicators to generate health status intervention feature data;

[0121] In this embodiment of the invention, an analysis of the health status of beekeeping behaviors and environmental indicators is conducted based on the correlation data between beekeeping health status and environment. The core logic is to screen beekeeping behaviors and environmental indicators that can be changed through human intervention and have a significant impact on health status, thus clarifying the boundary conditions for intervention. The criteria for defining interventionable features are: first, the indicator corresponding to the feature has a clear means of human intervention; second, the absolute value of the correlation strength between the feature and the health status is ≥0.6; and third, the range of regulation of the feature is within the feasible range of actual beekeeping operations. Based on these criteria, interventionable environmental indicators are screened from the correlation data between beekeeping health status and environment, including hive temperature, hive humidity, humidity in the core foraging area, and ammonia concentration in the hive (which can be controlled through ventilation, heating, humidification, and hive cleaning), while excluding non-interventional oxygen concentration and carbon dioxide concentration (which are greatly affected by the natural environment and whose human intervention costs are too high). Simultaneously, interventionable beekeeping behavior characteristics were screened from the health status-related feature data of beekeeping, including honey-foraging event density (which can be controlled by adjusting the planting range of nectar-producing plants), honey-making intensity stability (which can be controlled by adjusting the number of beehives), and flight event frequency (which can be controlled by optimizing the foraging path by adjusting the placement of beehives). The feasibility of controlling the screened interventionable characteristics was verified, and the control range and precision of each characteristic were determined. For example, the control range of beehive temperature was 20-35℃ with a control precision of 0.5℃; the control range of beehive humidity was 60%-85% with a control precision of 1%RH; the control range of beehive number was 8-12 frames with a control precision of 1 frame; and the control range of nectar-producing plant planting area was 30-50 meters around the beehive. For each modifiable feature, its optimal control target value (corresponding to the environmental indicator range or breeding behavior indicator range of the health level) and control threshold (exceeding this threshold results in an irreversible decline in health status) are defined. For example, the optimal target value for hive temperature is 25-30℃, and the control threshold is <22℃ or >33℃; the optimal target value for the number of bee combs is 10 frames, and the control threshold is <8 frames or >12 frames. The name, correlation strength, control range, control precision, optimal target value, and control threshold of the modifiable feature are integrated to generate health status modifiable feature data.

[0122] Step S44: Perform data segmentation processing on the health status interventionable feature data to conditionalize breeding behavior and the differences in environmental indicators, and generate breeding behavior conditionalization-environmental difference segmented data;

[0123] In this embodiment of the invention, data segmentation processing is performed on the health status interventionable feature data to categorize breeding behavior conditions and environmental indicator differences. The core logic is to divide the data into segments according to the type of breeding behavior conditions and the degree of difference in environmental indicators, so that each segment of data corresponds to a unified level of breeding behavior conditions and environmental differences, providing a refined data foundation for subsequent influence feature analysis. First, for the breeding behavior condition segmentation, different behavioral condition types are divided based on the interventionable breeding behavior characteristics. Then, for the environmental indicator difference segmentation, based on the distribution range of interventionable environmental indicators, five levels are divided according to the degree of difference: Level 1 corresponds to the optimal range of health (e.g., hive temperature 25-30℃), Level 2 corresponds to the edge range of health (e.g., 24-25℃ or 30-31℃), Level 3 corresponds to the sub-healthy range (e.g., 22-24℃ or 31-33℃), Level 4 corresponds to the edge range of abnormality (e.g., 20-22℃ or 33-35℃), and Level 5 corresponds to the severe range of abnormality (e.g., <20℃ or >35℃). The interval width of each level is equal to ensure the rationality of the difference division. Based on the above segmentation rules, the health status interventionable feature data are cross-segmented to form a combined data segment of "breeding behavior condition type - environmental indicator difference level", generating breeding behavior conditionalization - environmental difference segmented data.

[0124] Step S45: Based on the conditionalized breeding behavior-environmental difference segmented data and the bee breeding health status-environmental correlation data, conduct an environmental impact characteristic analysis of breeding health status and generate environmental impact characteristic data of breeding health status.

[0125] In this embodiment of the invention, environmental impact characteristic analysis of beekeeping health status is conducted based on segmented data of beekeeping behavior conditions and environmental differences, as well as beekeeping health status-environment correlation data. This quantifies the changing patterns of health status under different environmental difference levels and clarifies the influence weights and trends of environmental indicators on health status. Data segments with the same beekeeping behavior conditions but different environmental difference levels are extracted from the segmented data of beekeeping behavior conditions and environmental differences to construct an environmental impact analysis control group. For example, a fixed beekeeping behavior condition of "sufficient nectar source, 10 frames of beehives, and placement in a sunny location" is used to compare health status data under different temperature difference levels within the beehives. For each control group, the distribution ratio of health status levels is calculated. Simultaneously, the probability of a decrease in health status level for each change in environmental indicator level is calculated. Based on the correlation strength of beekeeping health status-environment correlation data, the influence weights of each environmental indicator are determined, and the weight values ​​are calculated using the analytic hierarchy process (AHP). Combining the influence weights and the probability of health status changes, an environmental impact characteristic model is constructed to clarify the critical influence thresholds and influence trend types (linear decline, non-linear decline, etc.) of each environmental indicator, generating environmental impact characteristic data of beekeeping health status.

[0126] Step S46: Based on the environmental impact characteristic data of beekeeping health status, conduct relevant characteristic analysis of health status and environmental indicators under different beekeeping behaviors to generate beekeeping health status-environment related characteristic data.

[0127] In this embodiment of the invention, based on the environmental impact characteristic data of beekeeping health status, a correlation characteristic analysis of health status and environmental indicators under different beekeeping behaviors is conducted. The differences in the correlation between environmental indicators and health status under different beekeeping behavior conditions are compared to clarify the optimal matching combination of beekeeping behaviors and environmental indicators. The data is grouped according to beekeeping behavior type, with each group containing relevant characteristic data for different levels of environmental differences. For example, it is divided into three groups: "Abundant Honey Source," "Moderate Honey Source," and "Scarce Honey Source." Each group covers environmental impact data for various temperature and humidity levels. For each group of data, the correlation coefficient between health status and each environmental indicator is calculated, and the changes in the correlation coefficient under different beekeeping behaviors are compared. For example, in the "Abundant Honey Source" group, the correlation coefficient between hive temperature and health status is -0.78, while in the "Scarce Honey Source" group, the correlation coefficient is -0.62, indicating that temperature has a more significant impact on health status when honey sources are abundant. Simultaneously, the percentage of healthy beehives corresponding to the optimal range of each environmental indicator under different beehive practices was statistically analyzed. For example, under the "10-frame beehive" practice, the percentage of healthy beehives at a temperature of 25-30℃ was 85%; under the "8-frame beehive" practice, the percentage of healthy beehives at the same temperature range was only 65%; and under the "12-frame beehive" practice, the percentage of healthy beehives was 70%, indicating that 10 frames of beehives is the optimal beehive practice matching the optimal temperature range. Based on the comparative analysis results, the optimal beehive practice combinations corresponding to the optimal range of each environmental indicator were selected. For example, when the beehive temperature is 25-30℃ and the humidity is 65%-75%, the optimal beehive practice combination is "sufficient nectar source, 10 frames of beehives, and placement in a sunny location," at which point the percentage of healthy beehives can reach 88%; when the ammonia concentration in the beehive is <0.001%, the optimal beehive practice combination is "regular cleaning of the beehive (once a week), 10 frames of beehives, and good ventilation," at which point the percentage of healthy beehives can reach 90%. By integrating correlation coefficients, health level percentages, and optimal matching combinations under different beekeeping behaviors, we can generate beekeeping health status-environment related characteristic data, providing a core basis for the formulation of subsequent beekeeping optimization management strategies.

[0128] Furthermore, step S45 includes the following steps:

[0129] By analyzing the gradient changes in beekeeping health status based on environmental indicators through segmented data of beekeeping behavior conditionalization and environmental differences, we can generate gradient change data of environmental differences in beekeeping health status. Furthermore, we can analyze the environmental impact characteristics of beekeeping health status based on the gradient change data of environmental differences in beekeeping health status, and generate environmental impact characteristic data of beekeeping health status.

[0130] In this embodiment of the invention, a gradient change analysis of beekeeping health status based on environmental index differences is conducted using segmented data of beekeeping behavior-environmental differences. A single combination of beekeeping behavior conditions is selected as a fixed variable from the segmented data of beekeeping behavior-environmental differences. The baseline beekeeping behavior condition is selected as "abundant nectar source (nectar source within 30-50 meters of the hive), 10 frames of honeycomb, and the hive placed in a sunny location (daily sunlight duration ≥ 8 hours)". All segmented data of environmental differences under this condition and the corresponding beekeeping health status-environmental correlation data are extracted. Based on the level of environmental index differences, a 5-level environmental gradient is constructed in order from level 1 to level 5. Each gradient corresponds to a unique range of environmental indicators. For example, the temperature gradient inside the hive is level 1 (25-30℃), level 2 (24-25℃ or 30-31℃), level 3 (22-24℃ or 31-33℃), level 4 (20-22℃ or 33-35℃), and level 5 (<20℃ or >35℃). For each environmental gradient, the corresponding health status level distribution data is statistically analyzed, and the proportion, mean, and standard deviation of each health status are calculated. The proportion of healthy status is calculated as the number of healthy status samples divided by the total number of samples in that gradient. The mean is the arithmetic mean of the comprehensive matching scores for health status within that gradient, and the standard deviation is an indicator of the dispersion of the comprehensive matching scores. Simultaneously, the changes in health status between adjacent gradients are calculated, including changes in the proportion of healthy status and changes in the comprehensive matching scores. These changes are calculated as the difference between the index values ​​of the subsequent gradient and the index values ​​of the previous gradient, thus characterizing the gradient change rate and generating environmental difference-aquaculture health status gradient change data.

[0131] An environmental impact characteristic analysis of the gradient change data of environmental differences and livestock health status was conducted, and the data was normalized. Based on the normalized data, the analytic hierarchy process (AHP) was used to determine the influence weights of each environmental indicator, constructing a hierarchical model. The target layer represents the degree of influence on livestock health status, the criterion layer represents each environmental indicator (beehive temperature, beehive humidity, humidity in the core foraging area, and ammonia concentration in the beehive), and the scenario layer represents the gradient change of each environmental indicator. The weight values ​​were calculated by pairwise comparison judgment matrices, and finally, the weights for beehive temperature (0.35), beehive humidity (0.25), core foraging area humidity (0.2), and beehive ammonia concentration (0.2) were determined. By combining gradient change data, critical characteristics of environmental impact are identified. Critical characteristics include critical impact threshold and impact trend type. The critical impact threshold is defined as the environmental gradient level corresponding to the first time the decline in the proportion of healthy individuals exceeds 20%. The impact trend type is determined by the distribution law of gradient change. If the absolute value of gradient change increases with the gradient level, it is a non-linear accelerating impact trend. If the absolute value of change remains stable, it is a linear impact trend. This generates environmental impact characteristic data of aquaculture health status.

[0132] Furthermore, step S5 includes the following steps:

[0133] Step S51: Analyze the adjustable beekeeping parameters based on the interventionable health status characteristic data, and generate adjustable beekeeping parameter data;

[0134] In this embodiment of the invention, the analysis of adjustable beekeeping parameters is carried out based on the interventionable health status feature data. The core logic is to decompose the interventionable features into specific quantifiable and human-controlled beekeeping parameters, and to clarify the control boundaries, precision, and correlation mapping relationship of each parameter with the health status. The interventionable health status feature data includes environmental interventionable features and beekeeping behavior interventionable features. Based on this, the types of adjustable beekeeping parameters are divided into two main categories: environmental control parameters and beekeeping behavior control parameters. The environmental control parameters correspond to the environmental interventionable characteristics, including hive temperature control parameters, hive humidity control parameters, core foraging area humidity control parameters, and hive ammonia concentration control parameters. Specifically, the hive temperature control parameter has a defined control range of 20-35℃, a control precision of 0.5℃, a minimum control step size of 0.5℃, and an optimal target range of 25-30℃; the hive humidity control parameter has a control range of 60%-85%RH, a control precision of 1%RH, a minimum control step size of 1%RH, and an optimal target range of 65%-75%RH; the core foraging area humidity control parameter has a control range of 60%-80%RH, a control precision of 2%RH, a minimum control step size of 2%RH, and an optimal target range of 65%-70%RH; and the hive ammonia concentration control parameter has a control range of 0-0.002%, a control precision of 0.0001%, and an optimal target value <0.001%. The parameters for regulating beekeeping behavior correspond to the interveneable characteristics of beekeeping behavior, including parameters for regulating the number of beehives, the range of nectar source cultivation, the location of beehives, and the beehive cleaning cycle. Specifically, the parameters for regulating the number of beehives range from 8 to 12 frames, with a control precision of 1 frame and an optimal target value of 10 frames; the parameters for controlling the range of nectar source cultivation range from 15 to 50 meters around the beehive, with a control precision of 5 meters and an optimal target range of 30 to 50 meters; the parameters for controlling the location of the beehives are quantified according to the duration of sunlight, ranging from 0 to more than 8 hours, with an optimal target value of ≥8 hours (sunny placement); and the parameters for controlling the beehive cleaning cycle range from 3 to 7 days / time, with an optimal target value of 7 days / time. The correlation and constraints between the adjustable parameters are also analyzed. For example, when the number of beehives increases by 1 frame, the optimal target range for temperature inside the beehive needs to be lowered by 0.5℃, and the optimal target range for humidity needs to be increased by 1%RH; when the nectar source cultivation range is expanded by 10 meters, the control precision for ammonia concentration inside the beehive needs to be increased to 0.00005%. The types, control ranges, control precision, minimum step size, optimal target values, and related constraints of all adjustable parameters are integrated to generate adjustable parameter data for beekeeping.

[0135] Step S52: Optimize the spatial design of beekeeping health status by using beekeeping health status-environment related characteristic data to generate beekeeping health status optimization spatial data, and perform global search optimization iteration processing on beekeeping health status optimization spatial data to generate optimized beekeeping health status data.

[0136] In this embodiment of the invention, an optimization space design for the health status of bees is carried out based on beekeeping health status-environment related characteristic data. The core logic is to define the parameter range that can be improved in the current health status based on the optimal matching combination of beekeeping behavior and environmental indicators, and to construct a multi-dimensional optimization space. Then, the optimal parameter combination is found through global search optimization iteration to generate optimized health status data. In the optimization space design stage, the adjustable parameters of beekeeping are used as dimensions, and the best matching combination in the beekeeping health status-environment related characteristic data is combined to determine the optimization range of each dimension, constructing a 5-dimensional optimization space (beehive temperature, beehive humidity, number of beehive combs, nectar source range, ammonia concentration). The optimization range of each dimension is ±20% of the optimal target range of the corresponding parameter. For example, the optimization range of beehive temperature is 20-36℃ (based on 25-30℃ expansion), and the optimization range of beehive number is 8-12 frames. At the same time, the health status and comprehensive matching score in the current beekeeping health status assessment data are used as the initial state values, and the optimization goals are defined as a comprehensive matching score ≥10 (healthy level) and a healthy level percentage ≥85%. In the global search optimization iteration phase, a grid search combined with gradient descent was used. The grid search phase divided search nodes according to the minimum adjustment step size of each parameter, generating a total of 20 × 26 × 5 × 8 × 20 = 416,000 search nodes. The gradient descent phase used the optimal node obtained from the grid search as the initial value, setting the learning rate to 0.05. The iteration termination condition was that the change in the comprehensive matching score over three consecutive iterations was <0.01 or the number of iterations reached 100. During each iteration, the parameter weights were adjusted based on the correlation coefficients in the beekeeping health status-environment related feature data, and the predicted health status score corresponding to each parameter combination was calculated. For example, when the initial parameter combination was "temperature 28℃, humidity 70%RH, 10 frames of bees, nectar source 30 meters, ammonia 0.001%", the score was 9.2. After iteratively adjusting the temperature to 29℃ and humidity to 72%RH, the score was 9.6. Further adjusting the nectar source to 40 meters resulted in a score of 10.1, achieving the optimization goal. After iteration, the optimal parameter combination and the corresponding predicted health status data are recorded, generating beekeeping health status optimization space data that includes the optimized parameter range, initial state value, optimization target value, optimal parameter combination, and predicted health status.

[0137] Step S53: Design an intelligent management strategy for optimized beekeeping based on optimized beekeeping health status data and adjustable beekeeping parameter data.

[0138] In this embodiment of the invention, an intelligent management strategy for optimized beekeeping is designed based on optimized beekeeping health status data and adjustable beekeeping parameter data. The core logic is to transform the optimized parameters into implementable control operation specifications, clarify the control trigger conditions, execution process, and monitoring feedback mechanism, forming a closed-loop management system. The strategy consists of three parts: an intelligent environmental control module, a precise control module for beekeeping behavior, and a dynamic monitoring and feedback module for health status. The intelligent environmental control module corresponds to adjustable environmental parameters, clearly defining the trigger thresholds and execution methods for each parameter: When the temperature inside the beehive is below 22℃, the electric heating device is activated, with the heating power increasing by 50W for every 1℃ decrease, until the temperature rises back to 25℃; when the temperature is above 33℃, the ventilation fan and misting cooling device are activated, with the fan speed at 3000r / min and the misting frequency at 5 minutes / time, until the temperature drops to 30℃; when the humidity inside the beehive is below 65%, the ultrasonic humidifier is activated, with a humidification rate of 10ml / h; when it is above 75%, the ventilation intensity is increased, and the fan speed is increased to 3500r / min; when the ammonia concentration inside the beehive is above 0.001%, the bottom ventilation vent and top exhaust fan of the beehive are activated, and a cleaning reminder is triggered to ensure that the beehive cleaning is completed within 72 hours. The precise control module for beekeeping behavior corresponds to adjustable parameters for beekeeping behavior, clearly defining the operation cycle and standards: the number of beehives is checked every 15 days; when the stability of honey production intensity is <80%, the number of beehives is adjusted to 10 frames; the nectar source planting area is assessed quarterly; when the honey harvesting event density is <5 times / square meter·hour, the nectar source planting area is expanded to 40-50 meters; the placement of beehives is adjusted annually in spring to ensure daily light exposure ≥8 hours. The dynamic monitoring and feedback module for health status clearly defines the monitoring frequency and feedback mechanism: environmental parameters are collected every 10 minutes, beekeeping behavior parameters are recorded daily, and health status assessments are conducted weekly; if the overall health matching score is <9.5, the parameter fine-tuning process is triggered, and the control threshold is corrected according to the parameter adjustment details in the optimized beekeeping health status data; if the score is <8, the emergency control process is initiated, increasing the monitoring frequency to 5 minutes / time, and simultaneously adjusting the parameters to the upper limit of the optimal range. The strategy also clarifies the device linkage logic of each module. For example, when the humidifier is started, the speed of the ventilation fan is reduced simultaneously, and the frequency of ammonia concentration monitoring is increased within 24 hours after cleaning the beehive, forming a closed-loop management of "monitoring-evaluation-control-feedback". The control specifications, triggering conditions, execution processes, and device parameters of each module are integrated to generate an intelligent management strategy for optimized beekeeping that includes strategy module division, operation standards, trigger thresholds, feedback mechanisms, and device linkage parameters.

[0139] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0140] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for analyzing and optimizing the health status of beekeeping based on artificial intelligence, characterized in that, Includes the following steps: Step S1: Use multi-source sensors to perform multi-source monitoring and sensing processing of the target bee area to generate multi-source monitoring and sensing data of the target bee area; Multi-dimensional hierarchical characteristic classification of the target bee area is performed by using multi-source monitoring and sensing data of the target bee area to generate multi-dimensional sensing characteristic data of the bee area. Step S2: Perform heterogeneous event combination feature analysis on the multidimensional perception characteristic data of bee regions to generate heterogeneous event combination feature data of bee behavior; Step S3: Perform beekeeping health status assessment processing on the combined feature data of heterogeneous bee behavior events to generate beekeeping health status assessment data; Step S4: Based on the multidimensional perception characteristic data of bee areas and the health status assessment data of bee farming, conduct relevant feature analysis on the health status and environmental indicators under different farming behaviors, and generate bee farming health status-environment related feature data. Step S5: Design intelligent management strategies for optimized beekeeping by using beekeeping health status and environmental-related characteristic data.

2. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 1, characterized in that, The multi-source monitoring and sensing data of the target bee area mentioned in step S1 includes bee behavior sensing data and bee area environmental data. The bee behavior sensing data includes bee flight image data, bee monitoring vibration signals, and bee monitoring voiceprint signals. The bee area environmental data includes area environmental temperature and humidity data and area environmental gas composition data.

3. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Design multi-source sensor topology configuration data; Step S12: Use the multi-source sensor topology configuration data to perform multi-source sensor topology configuration operation on the multi-source sensor, and use the configured multi-source sensor to perform multi-source bee monitoring and sensing processing on the target bee area to generate multi-source monitoring and sensing data of the target bee area. Step S13: Perform data consistency verification and adjustment processing on the multi-source monitoring and sensing data of the target bee area to generate standard multi-source monitoring and sensing data of the target bee area; Step S14: Perform time-series and spatial calibration processing on the multi-source monitoring and sensing data of the standard target bee area to generate spatiotemporal target bee area multi-source monitoring and sensing data; Step S15: Perform multi-dimensional hierarchical characteristic segmentation processing on the spatiotemporal target bee region multi-source monitoring and sensing data to generate bee region multi-dimensional sensing characteristic data.

4. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 3, characterized in that, Step S11 includes the following steps: Analyze bee activity correlation data in the target bee area to generate bee activity correlation data, wherein the bee activity correlation data includes beehive structure data and bee activity range data. Based on the correlation data of bee activity, the distribution characteristics of bee activity are analyzed to generate bee activity distribution characteristic data, and multi-source sensor topology configuration data is designed based on the bee activity distribution characteristic data.

5. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract bee behavior subdomain modalities based on bee region multidimensional perception characteristic data to generate bee behavior subdomain modal data; Step S22: Perform multimodal clustering feature analysis on the bee behavior subdomain modal data to generate multimodal clustering feature data on bee behavior, and perform bee behavior pattern analysis based on the multimodal clustering feature data on bee behavior to generate bee behavior pattern data; Step S23: Perform bee behavior event analysis on the bee behavior subdomain modal data to generate bee behavior event data; Step S24: Perform event combination difference analysis on bee behavior event data using bee behavior pattern data to generate bee behavior event combination difference data; Step S25: Perform heterogeneous event combination feature analysis on bee behavior based on the data of differences in bee behavior event combinations, and generate feature data of heterogeneous event combinations of bee behavior.

6. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 5, characterized in that, Step S23 includes the following steps: Step S231: Perform multimodal temporal feature analysis of bee behavior based on the bee behavior subdomain modal data, generate multimodal temporal feature data of bee behavior, and design a dynamic temporal window for bee behavior events based on the multimodal temporal feature data of bee behavior; Step S232: Perform bee behavior modality spatial feature analysis on the bee behavior subdomain modal data to generate bee behavior modality spatial feature data; Step S233: Detect bee behavior events using the dynamic temporal window of bee behavior events to generate bee behavior event data.

7. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform a correlation feature analysis on the health status of beekeeping based on the combined feature data of heterogeneous events in bee behavior, and generate correlation feature data on the health status of beekeeping. Step S32: Analyze the basic activity characteristics and operational behavior characteristics of bees based on the correlation characteristics data of bee health status; Step S33: Analyze the evolutionary trend of bee behavior based on the basic activity characteristics data and operational behavior characteristics data of bees, and generate bee behavior evolutionary trend data; Step S34: Obtain historical bee health status assessment data; Step S35: Design a multi-level discrimination matrix relationship for bee breeding health status based on historical bee health status assessment data, bee operational behavior data, and bee behavior evolution trend data, and generate a multi-level discrimination matrix for bee breeding health status. Step S36: Based on the multi-level discrimination matrix of bee breeding health status, perform bee breeding health status assessment processing on the combination feature data of heterogeneous events of bee behavior to generate bee breeding health status assessment data.

8. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Analyze the environmental distribution characteristics of bees based on the multidimensional perception characteristic data of bee regions, and generate bee environmental distribution characteristic data; Step S42: Perform health status and environmental correlation processing on the beekeeping health status assessment data and bee environmental distribution characteristic data to generate beekeeping health status-environment correlation data; Step S43: Based on the correlation data between beekeeping health status and environment, conduct health status intervention feature analysis on beekeeping behavior and environmental indicators to generate health status intervention feature data; Step S44: Perform data segmentation processing on the health status interventionable feature data to conditionalize breeding behavior and the differences in environmental indicators, and generate breeding behavior conditionalization-environmental difference segmented data; Step S45: Based on the conditionalized breeding behavior-environmental difference segmented data and the bee breeding health status-environmental correlation data, conduct an environmental impact characteristic analysis of breeding health status and generate environmental impact characteristic data of breeding health status. Step S46: Based on the environmental impact characteristic data of beekeeping health status, conduct relevant characteristic analysis of health status and environmental indicators under different beekeeping behaviors to generate beekeeping health status-environment related characteristic data.

9. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 8, characterized in that, Step S45 includes the following steps: By analyzing the gradient changes in beekeeping health status based on environmental indicators through segmented data of beekeeping behavior conditionalization and environmental differences, we can generate gradient change data of environmental differences in beekeeping health status. Furthermore, we can analyze the environmental impact characteristics of beekeeping health status based on the gradient change data of environmental differences in beekeeping health status, and generate environmental impact characteristic data of beekeeping health status.

10. The method for analyzing and optimizing the health status of beekeeping based on artificial intelligence according to claim 8, characterized in that, Step S5 includes the following steps: Step S51: Analyze the adjustable beekeeping parameters based on the interventionable health status characteristic data, and generate adjustable beekeeping parameter data; Step S52: Optimize the spatial design of beekeeping health status by using beekeeping health status-environment related characteristic data to generate beekeeping health status optimization spatial data, and perform global search optimization iteration processing on beekeeping health status optimization spatial data to generate optimized beekeeping health status data. Step S53: Design an intelligent management strategy for optimized beekeeping based on optimized beekeeping health status data and adjustable beekeeping parameter data.

Images