A modular intelligent control system and method for welfare poultry farming

By using a modular intelligent control system, which combines environmental parameters and poultry behavior data for comprehensive evaluation and closed-loop control, the system solves the problems of single environmental control methods and delayed early warning in existing technologies, and achieves precise identification of poultry behavior and early disease prevention and control.

CN123100532APending Publication Date: 2026-11-17SICHUAN TIANHUXINGNONG ECOLOGICAL AGRICULTURE TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611382065.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-08
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing smart farming systems suffer from data silos. Environmental control relies solely on sensor data for single-parameter closed-loop regulation, failing to incorporate the actual behavior and welfare levels of poultry into the control logic. This makes it difficult to meet the needs of early disease prevention and control. Furthermore, behavioral monitoring methods lack the ability to accurately identify time-series events, resulting in delayed early warnings.

Method used

A modular intelligent control system is adopted, including modules for data acquisition, intelligent environmental control, animal behavior analysis, welfare assessment, and cloud management. It acquires environmental parameters and video data through standardized communication protocols, uses a lightweight YOLO model and DeepSORT algorithm for poultry individual detection and trajectory tracking, and combines fuzzy PID algorithm and convolutional neural network for environmental adaptive adjustment and behavior recognition, generating a comprehensive welfare score and performing closed-loop control.

Benefits of technology

It enables precise environmental control based on poultry behavior, and can promptly identify and warn of abnormal behaviors such as lameness, aggression, and feather pecking, thereby improving the adaptability of the breeding environment and the level of health management, and meeting the needs of early disease prevention and control.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123100532A_ABST
    Figure CN123100532A_ABST
Patent Text Reader

Abstract

This invention relates to the field of poultry farming technology, specifically to a modular intelligent control system and method for welfare-oriented poultry farming. The system includes a data acquisition module, an intelligent environmental control module, an animal behavior analysis module, a welfare assessment module, and a cloud management module. The data acquisition module acquires environmental parameters and poultry behavior video data. The intelligent environmental control module generates environmental control instructions based on environmental parameters and welfare assessment results. The animal behavior analysis module identifies behaviors in the video data and outputs abnormal behavior warnings. The welfare assessment module generates a welfare score by combining environmental parameters and behavior identification results. The cloud management module coordinates data interaction and instruction issuance between the modules, forming a closed-loop control system. This invention solves the technical problem of existing environmental control methods being too simplistic and unable to meet the needs of early disease prevention and control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of poultry farming technology, specifically to a modular intelligent control system and method for the welfare farming of small poultry. Background Technology

[0002] With the increasing scale and intensification of livestock and poultry farming, small poultry farming has gradually shifted from traditional manual management to digital and intelligent management. The widespread use of various environmental sensors and video monitoring equipment in farms has made it possible to collect data on environmental parameters and poultry status. However, current smart farming systems generally suffer from data silos; subsystems such as environmental monitoring, video surveillance, and equipment control operate independently, lacking a unified data fusion and collaborative decision-making mechanism. This makes it difficult to transform the collected data into effective regulatory tools to improve animal welfare.

[0003] Specifically, this manifests in the following ways: 1) Environmental control relies solely on sensor data for single-parameter closed-loop regulation, without incorporating the actual behavior and welfare levels of poultry into the control logic, resulting in a crude control strategy; 2) Behavioral monitoring methods are mostly limited to activity level statistics, lacking the ability to precisely identify typical abnormal behaviors such as lameness, attacking and pecking feathers, and abnormal feeding, resulting in low identification accuracy, delayed early warning, and difficulty in meeting the needs of early disease prevention and control. Summary of the Invention

[0004] This invention provides a modular intelligent control system for welfare-oriented poultry farming, which solves the technical problem that the existing environmental control methods are too simple and cannot meet the needs of early disease prevention and control. The purpose of this invention is also to provide a modular intelligent control method for welfare-oriented poultry farming.

[0005] To address the aforementioned problems, the modular intelligent control system for small poultry welfare farming provided by this invention adopts the following technical solution: A modular intelligent control system for poultry welfare farming includes a data acquisition module, an intelligent environmental control module, an animal behavior analysis module, a welfare assessment module, and a cloud management module. The data acquisition module acquires environmental parameters and poultry behavior video data. The intelligent environmental control module generates environmental control commands based on the environmental parameters and welfare assessment results. The animal behavior analysis module performs behavior recognition on the video data and outputs abnormal behavior warnings. The welfare assessment module generates a welfare score by combining environmental parameters and behavior recognition results. The cloud management module coordinates data interaction and command issuance between the modules, forming a closed-loop control system.

[0006] Furthermore, the data acquisition module includes an environmental data access unit and a video data access unit; the environmental data access unit accesses real-time data from temperature, humidity, carbon dioxide, ammonia, and light intensity sensors through a standardized communication protocol, and performs noise reduction, missing value completion, and time-series alignment preprocessing on the raw data; the video data access unit is used to receive video streams acquired by visible light cameras and infrared cameras, extract frames at a fixed frame rate, and output them to the animal behavior analysis module.

[0007] Furthermore, the intelligent environmental control module includes a parameter threshold storage unit, an adaptive control algorithm unit, and an instruction generation unit; the parameter threshold storage unit is used to store differentiated environmental parameter thresholds according to poultry breeds and growth stages; the adaptive control algorithm unit adopts a fuzzy PID algorithm, taking the deviation between real-time environmental parameters and corresponding thresholds and the rate of change of deviation as input, and outputs the adjustment amounts of ventilation volume, humidification volume, and light parameters; the instruction generation unit converts the adjustment amounts into executable equipment control instructions and issues them.

[0008] Furthermore, the animal behavior analysis module includes a target detection unit, a trajectory tracking unit, and a temporal behavior recognition unit. The target detection unit uses a lightweight YOLO model to detect and locate poultry individuals in video frames. The trajectory tracking unit uses the DeepSORT algorithm to perform cross-frame association on the detected poultry individuals to generate continuous motion trajectories. The temporal behavior recognition unit uses a cascaded architecture of convolutional neural networks and long short-term memory networks to extract trajectory features and posture features, which are then input into the long short-term memory network for temporal modeling, outputting behavior categories and confidence levels.

[0009] Furthermore, the behavior categories output by the time-series behavior recognition unit include normal behavior and abnormal behavior. Abnormal behavior includes limping, attacking and pecking feathers, abnormal feeding, and huddling stress. When the confidence level of abnormal behavior exceeds a preset threshold and the duration exceeds a set time, an abnormal behavior warning signal is generated and pushed to the welfare assessment module and the cloud management module.

[0010] Furthermore, the welfare assessment module includes an indicator construction unit, a weight allocation unit, and a score calculation unit. The indicator construction unit extracts six assessment indicators from environmental parameters and behavior recognition results: temperature suitability, humidity suitability, air quality index, lighting suitability, activity richness, and abnormal behavior incidence rate. The weight allocation unit uses the analytic hierarchy process (AHP) to determine the weight of each indicator. The score calculation unit normalizes each indicator and then performs a weighted summation to output a comprehensive welfare score from 0 to 100.

[0011] Furthermore, the cloud management module includes a linkage decision-making unit, which has a preset mapping relationship between welfare scores and control strategies. When the overall welfare score is lower than a first threshold, a priority control strategy for environmental parameters is triggered. When the occurrence rate of abnormal behavior exceeds a second threshold, a behavior-oriented control strategy is triggered, including adjusting lighting parameters, adjusting ventilation volume, and sending a manual intervention prompt to the management terminal. The linkage decision-making unit records the changes in welfare scores before and after each control operation to continuously optimize the mapping relationship of control strategies.

[0012] The modular intelligent control system for poultry welfare farming provided by this invention has the following advantages: By using poultry behavior and welfare levels as references for environmental control, it changes the previous method of adjusting only single parameters based on sensor data, making the control strategy more aligned with the actual growth needs of poultry. Simultaneously, through multi-level image processing, it achieves individual poultry behavior recognition, accurately identifying and promptly warning of various typical abnormal behaviors such as lameness and feather pecking. This improves upon the traditional method of activity level statistics, which suffers from limited identification dimensions and untimely warnings, providing stable technical support for poultry health management and welfare improvement. Furthermore, it solves the technical problem of existing environmental control methods being too simplistic and failing to meet the needs of early disease prevention and control.

[0013] To address the above problems, the modular intelligent control method for welfare-oriented poultry farming provided by this invention adopts the following technical solution: A modular, intelligent, and regulated method for welfare-oriented poultry farming includes the following steps: S1. Data Acquisition: Real-time acquisition of breeding environment parameters through standardized communication protocols, and acquisition of poultry behavior video data at a fixed frame rate; S2. Environmental Adaptive Control: The real-time environmental parameters are compared with the preset differential thresholds according to the growth stage. The fuzzy PID algorithm is used to calculate the adjustment of ventilation volume, humidification volume and light parameters, and generate and issue control commands. S3. Behavior recognition and analysis: Individual poultry detection and cross-frame trajectory tracking are performed on video frames to extract motion trajectory features and posture features. A convolutional neural network and long short-term memory network cascaded model is used to perform temporal behavior recognition, determine whether there is abnormal behavior and generate early warning signals. S4. Comprehensive Welfare Assessment: Extract multiple assessment indicators from environmental parameters and behavioral identification results, assign weights using the analytic hierarchy process, calculate the weighted average, and output a comprehensive welfare score. S5. Closed-loop linkage decision-making: Based on the comprehensive welfare score and abnormal behavior warning signals, select and implement control strategies according to the preset mapping relationship, and record the control effects to iteratively optimize the strategies.

[0014] Furthermore, in step S3, the construction process of the cascaded model of the convolutional neural network and the long short-term memory network includes: collecting a dataset of poultry behavior videos labeled with behavioral categories, and expanding the sample size through data augmentation; using the convolutional neural network to extract the spatial features of each frame of the image, and inputting the spatial features of multiple consecutive frames into the long short-term memory network in chronological order for temporal dependency modeling; using the cross-entropy loss function for end-to-end training, and achieving lightweight deployment through model pruning and quantization operations.

[0015] Furthermore, in step S4, the activity richness index is calculated by weighting three sub-indicators: the total displacement of the individual poultry movement trajectory, the coverage area of ​​the activity area, and the diversity of behavioral categories; the abnormal behavior occurrence rate is the ratio of the number of abnormal behaviors identified per unit time to the total number of poultry; in step S5, when the occurrence rate of attack and feather pecking behavior exceeds a set threshold, the control strategy of reducing light intensity and adjusting spectrum is preferentially implemented.

[0016] The modular intelligent control method for poultry welfare farming provided by this invention has the following advantages: By using poultry behavior and welfare levels as references for environmental control, it changes the previous method of adjusting only single parameters based on sensor data, making the control strategy more aligned with the actual growth needs of poultry; at the same time, by using multi-level image processing to identify individual poultry behavior, it can accurately judge and promptly warn of various typical abnormal behaviors such as lameness and attacking feather pecking, improving the situation of traditional activity level statistics having a single identification dimension and untimely warnings, and providing stable technical support for poultry health management and welfare improvement; it also solves the technical problem of existing technologies having a single environmental control method that is difficult to meet the needs of early disease prevention and control. Attached Figure Description

[0017] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is an overall architecture diagram of the modular intelligent control poultry welfare breeding system provided by the present invention; Figure 2 This is an overall flowchart of the modular intelligent control poultry welfare breeding system provided by the present invention; Figure 3 This is a flowchart of the animal behavior recognition algorithm in this invention; Figure 4 This is a logic diagram of welfare assessment and closed-loop linkage in this invention. Detailed Implementation

[0018] The principles and spirit of the present invention will be explained in detail below with reference to several representative embodiments.

[0019] An embodiment of the modular intelligent control poultry welfare farming system provided by this invention: like Figure 1 As shown, the modular intelligent control system for small poultry welfare farming provided in this embodiment consists of a data acquisition module, an intelligent environmental control module, an animal behavior analysis module, a welfare assessment module, and a cloud management module. The modules interact and transmit commands through standardized communication links, forming a complete operating system centered on environmental perception, poultry status analysis, welfare level quantification, and intelligent equipment control. This system is adaptable to different breeds and growth stages of small livestock and poultry farming scenarios.

[0020] The data acquisition module is divided into an environmental data access unit and a video data access unit. The environmental data access unit receives real-time data from sensors for temperature, humidity, carbon dioxide, ammonia, and light intensity via a standardized communication protocol. It performs noise reduction, missing value completion, and time-series alignment preprocessing on the raw data to eliminate noise and time-series bias, providing a reliable data foundation for subsequent analysis. The video data access unit receives two video streams from a visible light camera and an infrared camera, covering image acquisition needs throughout the day and night. After the video streams are received, they are processed by frame extraction at a fixed frame rate before outputting the image frames to the animal behavior analysis module.

[0021] The intelligent environmental control module comprises a parameter threshold storage unit, an adaptive control algorithm unit, and an instruction generation unit. The parameter threshold storage unit stores differentiated environmental parameter thresholds categorized by poultry breed and growth stage, automatically matching the corresponding threshold benchmark based on the current batch. The adaptive control algorithm unit employs a fuzzy PID algorithm, taking the deviation and rate of change of real-time environmental parameters from their corresponding thresholds as input. After fuzzy rule inference and solution, it outputs the adjustment amounts for ventilation, humidification, and lighting parameters. The instruction generation unit receives the adjustment amounts, converts them into control instructions executable by the poultry equipment, and sends them to the corresponding execution terminals.

[0022] like Figure 3As shown, the animal behavior analysis module consists of a target detection unit, a trajectory tracking unit, and a temporal behavior recognition unit cascaded together. The target detection unit uses a lightweight YOLO model to process the input video frames, detecting and locating individual poultry within each frame, and outputting the bounding box and detection confidence score for each individual. The trajectory tracking unit uses the DeepSORT algorithm to perform cross-frame identity association on detected poultry individuals in consecutive frames, generating continuous motion trajectories for each poultry. The temporal behavior recognition unit employs a cascaded architecture of a convolutional neural network and a long short-term memory network. It first extracts the pose features of a single frame image and fuses them with the trajectory features. Then, it inputs the fused features from multiple consecutive frames into the long short-term memory network in chronological order for temporal dependency modeling, finally outputting the behavior category and corresponding confidence score.

[0023] The behavior categories output by the temporal behavior recognition unit include two types: normal behavior and abnormal behavior. Abnormal behavior includes four typical types: limping, aggressive feather pecking, abnormal feeding, and huddling stress. When the confidence level of a certain type of abnormal behavior exceeds a preset threshold and the duration exceeds a set time, the unit automatically generates an abnormal behavior warning signal and pushes it to the welfare assessment module and the cloud management module simultaneously.

[0024] like Figure 4 As shown, the welfare assessment module includes an indicator construction unit, a weight allocation unit, and a score calculation unit. The indicator construction unit extracts six assessment indicators from environmental parameters and behavior identification results: temperature suitability, humidity suitability, air quality index, lighting suitability, activity richness, and abnormal behavior incidence, covering two major assessment dimensions: environmental conditions and animal status. The weight allocation unit uses the analytic hierarchy process (AHP) to determine the weight percentage of each indicator; the weight configuration can be flexibly adjusted according to the breed and management objectives. The score calculation unit normalizes the raw data of each indicator, then sums them according to their corresponding weights, ultimately outputting a comprehensive welfare score from 0 to 100.

[0025] The cloud management module has a built-in linkage decision-making unit, which mainly undertakes the functions of data interaction and coordination between modules, unified distribution of instructions, and iterative optimization of control strategies. The linkage decision-making unit has a preset mapping relationship between welfare scores and control strategies: when the overall welfare score is lower than the first threshold, a priority control strategy for environmental parameters is triggered, prioritizing adjustments to environmental indicators such as temperature, humidity, ventilation, and lighting; when the incidence of abnormal behavior exceeds the second threshold, a behavior-oriented control strategy is triggered, specifically adjusting lighting parameters and ventilation volume, while simultaneously sending a manual intervention prompt to the management end. The linkage decision-making unit automatically records the changes in welfare scores before and after each control operation, continuously optimizing the mapping relationship of control strategies based on historical control data, and gradually improving the accuracy and adaptability of control.

[0026] In the modular intelligent control poultry welfare breeding system provided by this invention, during overall operation, the data acquisition module continuously collects breeding environment parameters and poultry behavior video data, which are then transmitted to the environmental intelligent control module and the animal behavior analysis module, respectively. The environmental intelligent control module completes basic environmental adaptive adjustment based on real-time environmental parameters, while the animal behavior analysis module simultaneously conducts behavior recognition and abnormal behavior warnings for individual poultry. The welfare assessment module calculates and generates a comprehensive welfare score based on environmental parameters and behavior recognition results, and feeds the score results back to the cloud management module. The cloud management module matches corresponding control strategies based on the comprehensive welfare score and abnormal warning information, and sends them down to the environmental intelligent control module for execution. At the same time, it retains the effect data before and after control for subsequent strategy iterations, thereby forming a complete closed-loop control mechanism of data acquisition, status analysis, welfare assessment, command control, and strategy optimization.

[0027] An embodiment of the modular intelligent control method for welfare-oriented poultry farming provided by this invention: like Figure 2 As shown, a modular intelligent control method for welfare-oriented poultry farming includes the following steps: S1. Data Acquisition: Real-time acquisition of breeding environment parameters through standardized communication protocols, and acquisition of poultry behavior video data at a fixed frame rate; S2. Environmental Adaptive Control: The real-time environmental parameters are compared with the preset differential thresholds according to the growth stage. The fuzzy PID algorithm is used to calculate the adjustment of ventilation volume, humidification volume and light parameters, and generate and issue control commands. S3. Behavior recognition and analysis: Individual poultry detection and cross-frame trajectory tracking are performed on video frames to extract motion trajectory features and posture features. A convolutional neural network and long short-term memory network cascaded model is used to perform temporal behavior recognition, determine whether there is abnormal behavior and generate early warning signals. S4. Comprehensive Welfare Assessment: Extract multiple assessment indicators from environmental parameters and behavioral identification results, assign weights using the analytic hierarchy process, calculate the weighted average, and output a comprehensive welfare score. S5. Closed-loop linkage decision-making: Based on the comprehensive welfare score and abnormal behavior warning signals, select and implement control strategies according to the preset mapping relationship, and record the control effects to iteratively optimize the strategies.

[0028] Specifically, in step S3, the construction process of the cascaded model of the convolutional neural network and the long short-term memory network includes: collecting a dataset of poultry behavior videos labeled with behavioral categories, and expanding the sample size through data augmentation; using the convolutional neural network to extract the spatial features of each frame of the image, and inputting the spatial features of multiple consecutive frames into the long short-term memory network in chronological order for temporal dependency modeling; using the cross-entropy loss function for end-to-end training, and achieving lightweight deployment through model pruning and quantization operations.

[0029] In step S4, the activity richness index is calculated by weighting three sub-indicators: the total displacement of the individual poultry movement trajectory, the coverage area of ​​the activity area, and the diversity of behavioral categories; the abnormal behavior occurrence rate is the ratio of the number of abnormal behaviors identified per unit time to the total number of poultry; in step S5, when the occurrence rate of attack and feather pecking behavior exceeds a set threshold, the control strategy of reducing light intensity and adjusting spectrum is implemented first.

[0030] Based on the above description in this specification, those skilled in the art will also understand that the following terms, such as "upper," "lower," "front," "rear," "left," "right," "width," "horizontal," "top," "bottom," "inner," and "outer," which indicate orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings of this specification. They are only for the purpose of facilitating the explanation of the present invention and simplifying the description, and do not explicitly or implicitly suggest that the device or element involved must have the specific orientation, or be constructed and operated in a specific orientation. Therefore, the above-mentioned orientation or positional relationship terms should not be understood or interpreted as limitations on the present invention.

[0031] In addition, in the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

Claims

1. A modular intelligent control system for the welfare farming of small poultry, characterized in that: It includes a data acquisition module, an intelligent environmental control module, an animal behavior analysis module, a welfare assessment module, and a cloud management module; the data acquisition module is used to acquire breeding environment parameters and poultry behavior video data; the intelligent environmental control module is used to generate environmental control instructions based on environmental parameters and welfare assessment results; the animal behavior analysis module is used to perform behavior recognition on video data and output abnormal behavior warnings. The welfare assessment module is used to generate welfare scores by combining environmental parameters and behavior recognition results; the cloud management module is used to coordinate data interaction and command issuance between modules to form a closed-loop control.

2. The modular intelligent control system for small poultry welfare farming according to claim 1, characterized in that: The data acquisition module includes an environmental data access unit and a video data access unit. The environmental data access unit accesses real-time data from temperature, humidity, carbon dioxide, ammonia and light intensity sensors through a standardized communication protocol, and performs noise reduction, missing value completion and time-series alignment preprocessing on the raw data. The video data access unit is used to receive video streams collected by visible light cameras and infrared cameras, extract frames at a fixed frame rate, and output them to the animal behavior analysis module.

3. The modular intelligent control system for small poultry welfare farming according to claim 1, characterized in that: The intelligent environmental control module includes a parameter threshold storage unit, an adaptive control algorithm unit, and an instruction generation unit. The parameter threshold storage unit is used to store differentiated environmental parameter thresholds according to poultry breeds and growth stages. The adaptive control algorithm unit adopts a fuzzy PID algorithm, taking the deviation between real-time environmental parameters and corresponding thresholds and the rate of change of the deviation as input, and outputting the adjustment amounts of ventilation volume, humidification volume, and light parameters. The instruction generation unit converts the adjustment amounts into executable equipment control instructions and issues them.

4. The modular intelligent control system for small poultry welfare farming according to any one of claims 1 to 3, characterized in that: The animal behavior analysis module includes a target detection unit, a trajectory tracking unit, and a temporal behavior recognition unit. The target detection unit uses a lightweight YOLO model to detect and locate poultry individuals in video frames. The trajectory tracking unit uses the DeepSORT algorithm to perform cross-frame association on the detected poultry individuals and generate continuous motion trajectories. The temporal behavior recognition unit uses a cascaded architecture of convolutional neural networks and long short-term memory networks to extract trajectory features and posture features, which are then input into the long short-term memory network for temporal modeling, outputting behavior categories and confidence levels.

5. The modular intelligent control system for small poultry welfare farming according to claim 4, characterized in that: The behavior categories output by the time-series behavior recognition unit include normal behavior and abnormal behavior. Abnormal behavior includes limping, attacking and pecking feathers, abnormal feeding, and huddling stress. When the confidence level of abnormal behavior exceeds a preset threshold and the duration exceeds a set time, an abnormal behavior warning signal is generated and pushed to the welfare assessment module and the cloud management module.

6. The modular intelligent control system for small poultry welfare farming according to any one of claims 1 to 3, characterized in that: The welfare assessment module includes an indicator construction unit, a weight allocation unit, and a score calculation unit. The indicator construction unit extracts six assessment indicators from environmental parameters and behavior recognition results: temperature suitability, humidity suitability, air quality index, lighting suitability, activity richness, and abnormal behavior incidence rate. The weight allocation unit uses the analytic hierarchy process (AHP) to determine the weight of each indicator. The score calculation unit normalizes each indicator and then performs a weighted summation to output a comprehensive welfare score from 0 to 100.

7. The modular intelligent control system for small poultry welfare farming according to claim 6, characterized in that: The cloud management module includes a linkage decision-making unit, which has a preset mapping relationship between welfare scores and control strategies. When the overall welfare score is lower than a first threshold, a priority control strategy for environmental parameters is triggered. When the occurrence rate of abnormal behavior exceeds a second threshold, a behavior-oriented control strategy is triggered, including adjusting lighting parameters, adjusting ventilation volume, and sending a manual intervention prompt to the management terminal. The linkage decision-making unit records the changes in welfare scores before and after each control operation to continuously optimize the mapping relationship of control strategies.

8. A modular, intelligently regulated method for welfare-oriented poultry farming, characterized in that, Includes the following steps: S1. Data Acquisition: Real-time acquisition of breeding environment parameters through standardized communication protocols, and acquisition of poultry behavior video data at a fixed frame rate; S2. Environmental Adaptive Control: The real-time environmental parameters are compared with the preset differential thresholds according to the growth stage. The fuzzy PID algorithm is used to calculate the adjustment of ventilation volume, humidification volume and light parameters, and generate and issue control commands. S3. Behavior recognition and analysis: Individual poultry detection and cross-frame trajectory tracking are performed on video frames to extract motion trajectory features and posture features. A convolutional neural network and long short-term memory network cascaded model is used to perform temporal behavior recognition, determine whether there is abnormal behavior and generate early warning signals. S4. Comprehensive Welfare Assessment: Extract multiple assessment indicators from environmental parameters and behavioral identification results, assign weights using the analytic hierarchy process, calculate the weighted average, and output a comprehensive welfare score. S5. Closed-loop linkage decision-making: Based on the comprehensive welfare score and abnormal behavior warning signals, select and implement control strategies according to the preset mapping relationship, and record the control effects to iteratively optimize the strategies.

9. The modular intelligent control method for welfare-oriented poultry farming according to claim 8, characterized in that: In step S3, the construction process of the cascaded model of the convolutional neural network and the long short-term memory network includes: collecting a dataset of poultry behavior videos labeled with behavioral categories, and expanding the sample size through data augmentation; using the convolutional neural network to extract the spatial features of each frame of the image, and inputting the spatial features of multiple consecutive frames into the long short-term memory network in chronological order for temporal dependency modeling; using the cross-entropy loss function for end-to-end training, and achieving lightweight deployment through model pruning and quantization operations.

10. The modular intelligent control method for welfare-oriented poultry farming according to claim 8, characterized in that: In step S4, the activity richness index is calculated by weighting three sub-indicators: the total displacement of the individual poultry movement trajectory, the coverage area of ​​the activity area, and the diversity of behavioral categories; the abnormal behavior occurrence rate is the ratio of the number of abnormal behaviors identified per unit time to the total number of poultry; in step S5, when the occurrence rate of attack and feather pecking behavior exceeds a set threshold, the control strategy of reducing light intensity and adjusting spectrum is implemented first.