Breeding environment intelligent regulation and control method, system and equipment based on Internet of Things

Abstract

The invention discloses a breeding environment intelligent regulation and control method, system and device based on the Internet of Things, and relates to the technical field of intelligent regulation and control, and the method comprises the steps: carrying out the distributed deployment of a multi-mode environment sensing array; constructing a global multi-modal environment dynamic heat map; the method comprises the following steps: synchronously acquiring an infrared thermal imaging data stream and a visible light video data stream, fusing a global multi-modal environment dynamic heat map to perform coupling clustering analysis, and dividing a chicken house plane into a plurality of dynamic comprehensive regulation and control partitions; setting a plurality of environment adjustment optimization targets to carry out environment parameter adjustment rolling optimization, and outputting a plurality of time sequence control parameter groups; conflict detection fusion of the multiple sequential control parameter groups is carried out, a linkage control strategy is output, and intelligent regulation and control of the breeding environment of the henhouse are executed. The technical problems that in the prior art, the spatial distribution state of the breeding environment cannot be accurately sensed, the regulation and control strategy is extensive, and the actual requirements of chicken flocks are difficult to dynamically match are solved, and the technical effect of achieving refined, dynamic and intelligent regulation and control of the breeding environment is achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to intelligent control methods, systems, and equipment for aquaculture environments based on the Internet of Things. Background Technology

[0002] In large-scale poultry farming, the internal environment of chicken houses is affected by factors such as ventilation structure, equipment layout, and flock activity, resulting in a significant uneven distribution of parameters such as temperature, humidity, harmful gases, and dust. Existing farming management methods rely heavily on a small number of fixed monitoring points to obtain environmental data, which is insufficient to reflect the true environmental conditions of the chicken house as a whole or in specific areas, leading to a lack of detailed description of environmental spatial distribution information.

[0003] Meanwhile, environmental control typically relies on unified control of the entire chicken house, with adjustments primarily based on threshold judgments of environmental parameters. This lacks comprehensive analysis of the flock's physical condition and behavioral changes, making it difficult to reflect the flock's physiological and behavioral responses to environmental changes in real time. When there are significant environmental differences between different areas of the chicken house or when the flock distribution is dynamically changing, these control methods are unable to dynamically adjust according to the actual needs of the flock, easily leading to over- or under-regulation in localized areas and affecting the stability of the farming operation. Summary of the Invention

[0004] This application provides a method, system, and equipment for intelligent control of the breeding environment based on the Internet of Things, which is used to address the technical problems in the prior art where the spatial distribution of the breeding environment cannot be accurately perceived, the control strategies are extensive, and it is difficult to dynamically match the actual needs of the chicken flock.

[0005] In view of the above problems, this application provides a method, system and equipment for intelligent control of aquaculture environment based on Internet of Things.

[0006] The first aspect of this application provides a method for intelligent control of aquaculture environment based on the Internet of Things, the method comprising: A distributed deployment of a multimodal environmental sensing array is carried out in the chicken house based on a preset radius scale. Microbial concentration reference information returned from the laboratory is received as a baseline truth value. The multimodal environmental sensing data returned by the multimodal environmental sensing array is aligned with this data, and spatial interpolation prediction based on fluid dynamics simulation is performed to construct a global multimodal environmental dynamic heat map. Infrared thermal imaging data streams and visible light video data streams are simultaneously collected by deploying thermal infrared cameras and visible light cameras covering the chicken house. The global multimodal environmental dynamic heat map is fused for coupled cluster analysis, dividing the chicken house plane into multiple dynamic integrated control zones. Multiple environmental regulation optimization targets are set based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones. Environmental parameter tuning is performed through rolling optimization, outputting multiple time-series control parameter groups. Based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, conflict detection and fusion of the multiple time-series control parameter groups are performed, outputting a linkage control strategy to execute intelligent control of the chicken house's breeding environment.

[0007] A second aspect of this application provides an intelligent control system for aquaculture environments based on the Internet of Things, the system comprising: The system comprises the following modules: a deployment module for distributed deployment of a multimodal environmental sensing array within the chicken house based on a preset radius scale; a prediction module for receiving microbial concentration reference information from the laboratory as a baseline, aligning it with the multimodal environmental sensing data returned by the multimodal environmental sensing array, performing spatial interpolation prediction based on fluid dynamics simulation, and constructing a global multimodal environmental dynamic heat map; a clustering analysis module for simultaneously acquiring infrared thermal imaging data streams and visible light video data streams by deploying thermal infrared cameras and visible light cameras covering the chicken house, fusing the global multimodal environmental dynamic heat map for coupled clustering analysis, and dividing the chicken house plane into multiple dynamic integrated control zones; a rolling optimization module for setting multiple environmental regulation optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, performing rolling optimization of environmental parameters, and outputting multiple time-series control parameter groups; and an intelligent control module for performing conflict detection and fusion of the multiple time-series control parameter groups based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, outputting a linkage control strategy, and executing intelligent control of the chicken house's breeding environment.

[0008] A second aspect of this application provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the IoT-based intelligent control method for aquaculture environment provided in this application.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application deploys a distributed multimodal environmental sensing array in a chicken house based on a preset radius scale; it receives microbial concentration reference information returned from the laboratory as a baseline truth value, aligns it with the multimodal environmental sensing data returned from the multimodal environmental sensing array, performs spatial interpolation prediction based on fluid dynamics simulation, and constructs a global multimodal environmental dynamic heat map; it deploys thermal infrared cameras and visible light cameras covering the chicken house to simultaneously collect infrared thermal imaging data streams and visible light video data streams, fuses the global multimodal environmental dynamic heat map for coupled cluster analysis, and divides the chicken house plane into multiple dynamic integrated control zones; it sets multiple environmental regulation optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, performs rolling optimization of environmental parameters, and outputs multiple time-series control parameter groups; based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, it performs conflict detection and fusion of the multiple time-series control parameter groups, outputs a linkage control strategy, and executes intelligent control of the chicken house's breeding environment. This invention addresses the technical problems in existing technologies, such as the inability to accurately perceive the spatial distribution of the breeding environment, the coarse control strategies, and the difficulty in dynamically matching the actual needs of chicken flocks. By constructing a global multimodal dynamic heat map of the environment and performing dynamic partitioning and rolling optimization control accordingly, the invention achieves the technical effect of refined, dynamic, and intelligent control of the breeding environment. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A schematic diagram of the process of an IoT-based intelligent control method for aquaculture environment provided in an embodiment of this application; Figure 2 A schematic diagram of the structure of an IoT-based intelligent aquaculture environment control system provided in this application embodiment; Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application.

[0012] Explanation of reference numerals in the attached diagram: Deployment module 11, Prediction module 12, Cluster analysis module 13, Scrolling optimization module 14, Intelligent control module 15, Bus 300, Receiver 301, Processor 302, Transmitter 303, Memory 304, Bus interface 305. Detailed Implementation

[0013] This application provides an IoT-based intelligent control method, system, and equipment for the breeding environment. It addresses the technical problems in existing technologies, such as the inability to accurately perceive the spatial distribution of the breeding environment, the coarse control strategies, and the difficulty in dynamically matching the actual needs of the chicken flock. By constructing a global multimodal dynamic heat map of the environment and performing dynamic partitioning and rolling optimization control based on it, the technical effect of achieving refined, dynamic, and intelligent control of the breeding environment is achieved.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0015] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0016] Example 1, as Figure 1 As shown, this application provides an intelligent control method for aquaculture environment based on the Internet of Things, the method comprising: Step S100: Distribute multimodal environmental sensor arrays in the chicken coop based on a preset radius scale.

[0017] In this embodiment, the preset radius scale is a pre-defined distance parameter used to limit the spatial range covered by each environmental monitoring point. When deploying a distributed multimodal environmental sensor array in a chicken house based on the preset radius scale, the preset radius scale is first used as the basic constraint for environmental monitoring coverage, according to the overall size and spatial structure of the chicken house. Then, multiple environmental monitoring point locations are selected inside the chicken house according to the preset radius scale, ensuring that the spacing between adjacent environmental monitoring points meets the coverage requirements of the preset radius scale, thereby achieving continuous environmental coverage of the chicken house space. After determining the monitoring point locations, environmental sensing devices capable of simultaneously acquiring multiple environmental parameters are installed at each environmental monitoring point, making each environmental monitoring point a multimodal environmental sensing node. Multiple multimodal environmental sensing nodes together constitute a multimodal environmental sensor array in a distributed manner within the chicken house.

[0018] Step S200: Receive the microbial concentration reference information returned from the laboratory as the baseline truth value, align with the multimodal environmental sensing data returned from the multimodal environmental sensing array, perform spatial interpolation prediction based on fluid dynamics simulation, and construct a global multimodal environmental dynamic heat map.

[0019] In this embodiment, microbial concentration reference information is first received from the laboratory, which is high-accuracy microbial concentration data obtained through laboratory testing. Then, the multimodal environmental sensing data transmitted from the multimodal environmental sensing array is spatiotemporally aligned, and the aligned multimodal environmental sensing data is input into a livestock environment fluid dynamics simulation model to perform transient fluid field solving, obtaining the temperature field, humidity field, harmful gas concentration field, and dust concentration field inside the chicken house. Subsequently, a risk probability model is constructed based on the statistical correlation between microbial concentration and environmental parameters in the chicken house. The temperature field, humidity field, harmful gas concentration field, and dust concentration field, along with the microbial concentration reference information, are then incorporated into the risk probability model to extrapolate microbial risk probability. Finally, a microbial risk prediction field characterizing the risk of microbial distribution in the chicken house is generated, thereby constructing a global multimodal environmental dynamic heat map.

[0020] Furthermore, the method provided in the application embodiments also includes: Each multimodal environmental sensing node in the multimodal environmental sensing array supports real-time sensing of temperature, humidity, ammonia, hydrogen sulfide, carbon dioxide, and dust concentration. The global multimodal environmental dynamic heat map includes a temperature field, a humidity field, a harmful gas concentration field, a dust concentration field, and a microbial risk prediction field.

[0021] In this embodiment, the multimodal environmental sensing array consists of multiple multimodal environmental sensing nodes. Each multimodal environmental sensing node has the ability to continuously and in real time sense the temperature, humidity, ammonia, hydrogen sulfide, carbon dioxide and dust concentration in the chicken house environment, and is used to obtain multidimensional environmental parameter information at different spatial locations.

[0022] Based on environmental parameter data collected by various multimodal environmental sensing nodes at different locations within the chicken house, environmental information is spatially integrated and expressed to form a global multimodal environmental dynamic heat map that characterizes the overall environmental state of the chicken house. The global multimodal environmental dynamic heat map includes a temperature field reflecting the spatial temperature distribution, a humidity field reflecting the spatial humidity distribution, a harmful gas concentration field reflecting the spatial distribution of various harmful gases, a dust concentration field reflecting the spatial distribution of dust, and a microbial risk prediction field reflecting the risk of microbial distribution.

[0023] Furthermore, in the method provided in the application embodiment, receiving the microbial concentration reference information returned from the laboratory as the baseline truth value, aligning it with the multimodal environmental sensing data returned from the multimodal environmental sensing array, performing spatial interpolation prediction based on fluid dynamics simulation, and constructing a global multimodal environmental dynamic heat map, further includes: After spatiotemporal alignment of the multimodal environmental sensing data based on the multimodal environmental sensing array, the transient fluid field is solved by inputting it into the aquaculture environment fluid dynamics simulation model, and the temperature field, humidity field, harmful gas concentration field, and dust concentration field are output. Based on the statistical correlation between microbial concentration and environmental parameters in the chicken house, a risk probability model is constructed. Combining the temperature field, humidity field, harmful gas concentration field, and dust concentration field, the risk probability model is used to extrapolate the microbial risk probability of the microbial concentration reference information, and the microbial risk prediction field is generated.

[0024] In this embodiment, the multimodal environmental sensing data collected by each multimodal environmental sensing node in the multimodal environmental sensing array is first organized. The data collected by different nodes at different times are then processed synchronously according to a unified time stamp. Based on the fixed installation position of each sensing node in the chicken house, the corresponding environmental parameter data are mapped to a unified spatial coordinate position, thereby completing the spatiotemporal alignment of the multimodal environmental sensing data. This ensures that data such as temperature, humidity, harmful gas concentration, and dust concentration can correspond one-to-one at the same time point and a specific spatial location. For example, when multiple sensing nodes located in different areas of the chicken house collect environmental data in adjacent time periods, the spatiotemporal alignment process can unify these data into a description of the environmental state of different locations in the chicken house at the same time.

[0025] After completing spatiotemporal alignment, the processed multimodal environmental sensing data is input into the poultry farming environment fluid dynamics simulation model. Based on the chicken house's spatial structure and airflow characteristics, this model calculates the flow, diffusion, and transmission of air within the chicken house. By advancing the data time-by-time, it obtains the changes in the internal environmental state of the chicken house, thus achieving transient fluid field solutions. Through this process, the discrete environmental monitoring data, originally existing only at sensor node locations, is expanded into a continuous distribution covering the entire chicken house space. This results in the output of a temperature field reflecting the temperature distribution, a humidity field reflecting the humidity distribution, a harmful gas concentration field reflecting the distribution of ammonia, hydrogen sulfide, and carbon dioxide, and a dust concentration field reflecting the dust distribution.

[0026] After obtaining the temperature field, humidity field, harmful gas concentration field, and dust concentration field covering the chicken coop space, the reference information on microbial concentrations obtained from the laboratory at different times and sampling locations is first organized. The corresponding sampling time and sampling location are marked for each microbial concentration reference information. The temperature value, humidity value, harmful gas concentration value, and dust concentration value at the sampling time and sampling location are extracted from the temperature field, humidity field, harmful gas concentration field, and dust concentration field. This makes each microbial concentration reference information form a one-to-one correspondence with a set of environmental parameter values, thereby obtaining an environmental-microbial sample set for establishing statistical correlation laws.

[0027] Next, a risk probability model is constructed based on the statistical correlation between microbial concentration and environmental parameters in the chicken coop. In this process, the sample records in the environmental-microbial sample set are statistically summarized. Temperature, humidity, harmful gas concentration, and dust concentration are divided into several numerical intervals, and the distribution of microbial concentration reference information under different interval combinations is statistically analyzed. This yields the proportion of microbial concentration falling into different numerical ranges under given temperature, humidity, harmful gas, and dust concentration intervals. This proportion is then used as a probability mapping relationship and embedded in the risk probability model, enabling the model to accept temperature, humidity, harmful gas concentration, and dust concentration as inputs and output the corresponding microbial risk probability results.

[0028] Then, the microbial risk probability was extrapolated by combining the temperature field, humidity field, harmful gas concentration field, and dust concentration field with the microbial concentration reference information. In this process, using the spatial coordinates of the chicken coop as an index, the temperature, humidity, harmful gas concentration, and dust concentration values ​​were read point-by-point at each spatial location, and these four values ​​were fed as a set of inputs into the risk probability model. Next, during the calculation, the risk probability model first determined the corresponding interval combination conditions based on the interval to which each input value belonged. Then, it searched for the statistical distribution of the microbial concentration reference information under the interval combination conditions in the risk probability model, and finally converted this statistical distribution into the microbial risk probability output value for that spatial location. When the input value is near the boundary between two adjacent intervals, the risk probability model averaged the probability output values ​​of the adjacent interval combinations to ensure that the probability output remains continuous with changes in environmental parameters.

[0029] After completing point-by-point calculations for all spatial locations within the chicken house, the microbial risk probability output values ​​obtained from each spatial location are aggregated according to spatial coordinates and correspond one-to-one with the plane or three-dimensional spatial locations of the chicken house to form a continuously spatially distributed microbial risk prediction field.

[0030] Furthermore, the method provided in the application embodiments also includes: Based on the three-dimensional physical structure of the chicken house, the layout of the ventilation equipment, and the performance information of the ventilation equipment, a parametric fluid dynamics mesh model is established. Using the spatiotemporally aligned microbial concentration reference information and multimodal environmental perception data as joint constraints, the dynamic boundary parameters of the parametric fluid dynamics mesh model are calibrated in reverse through a data assimilation algorithm, and a simulation model of the aquaculture environment fluid dynamics is output.

[0031] In this embodiment, the geometric dimensions of the chicken house, such as length, width, and height, are first obtained based on the three-dimensional physical structure of the chicken house, and the location range of structures such as walls, floors, ceilings, and partitions in the chicken house space is determined. Then, the specific installation positions of fans, air inlets, and air outlets are determined based on the layout of ventilation equipment. Subsequently, the performance information of ventilation equipment is read to determine the air volume parameters and operating capacity parameters corresponding to each ventilation equipment. After completing the above information processing, the chicken house space is divided into regular grids at fixed intervals, and the structures such as walls and partitions are mapped as grid boundaries. At the same time, the air inlets and air outlets are mapped as the ventilation positions of the grid boundaries, and the air volume parameters are written into the grid boundary conditions as boundary input parameters, thereby establishing a parameterized fluid dynamics grid model that can use parameters to describe the chicken house space structure and ventilation conditions.

[0032] Subsequently, the microbial concentration reference information and multimodal environmental sensing data were spatiotemporally aligned to establish a one-to-one correspondence between the microbial concentration reference information and multimodal environmental sensing data such as temperature, humidity, harmful gas concentration, and dust concentration at the same time point and spatial location. Then, the spatiotemporally aligned microbial concentration reference information and multimodal environmental sensing data were used as joint constraint inputs to a parameterized fluid dynamics grid model. Under the current dynamic boundary parameter conditions, the parameterized fluid dynamics grid model was run to obtain the environmental parameter distribution results within the chicken coop space. The model calculation results were extracted at the spatial locations corresponding to the multimodal environmental sensing data and compared point-by-point with the actual collected multimodal environmental sensing data to calculate the differences between the model calculation results and the actual monitoring results.

[0033] After obtaining the difference results, the dynamic boundary parameters in the parametric fluid dynamics mesh model are adjusted in reverse to reduce the difference in the next model run. These dynamic boundary parameters include inlet supply air volume parameters, outlet exhaust air volume parameters, and boundary input parameters related to ventilation conditions. After completing one parameter adjustment, the parametric fluid dynamics mesh model is run again, and the difference calculation and parameter adjustment process is repeated until the temperature, humidity, harmful gas concentration, and dust concentration results output by the parametric fluid dynamics mesh model and the differences between these values ​​and the multimodal environmental sensing data at the corresponding time and spatial locations meet preset requirements, and the model's calculation results are consistent with the microbial concentration reference information.

[0034] After completing the above-mentioned reverse calibration process of dynamic boundary parameters, the calibrated parameterized fluid dynamics mesh model is output as the simulation model of aquaculture environment fluid dynamics, so that the simulation model of aquaculture environment fluid dynamics can perform stable and continuous calculations of air flow state and spatial distribution of environmental parameters under the actual ventilation operation conditions of chicken house.

[0035] Step S300: By deploying thermal infrared cameras and visible light cameras covering the chicken house to simultaneously collect infrared thermal imaging data streams and visible light video data streams, and fusing the global multimodal environmental dynamic heat map to perform coupled cluster analysis, the chicken house plane is divided into multiple dynamic comprehensive control zones.

[0036] In this embodiment, by deploying thermal infrared cameras and visible light cameras covering the entire space within the chicken coop, infrared thermal imaging data streams and visible light video data streams are simultaneously acquired. A deep visual recognition model is then used to fuse and analyze the two types of data, resulting in a chicken flock surface temperature distribution heatmap representing the spatial distribution of chicken flock surface temperature and a chicken flock behavior state vector representing the flock's behavioral characteristics. Subsequently, the chicken flock surface temperature distribution heatmap and chicken flock behavior state vector are fused with a global multimodal environmental dynamic heatmap to construct an environmental-biological coupling state matrix describing the coupling relationship between environmental factors and biological states. Finally, spatial clustering analysis is performed on the environmental-biological coupling state matrix. Based on the similarity of different regions in environmental parameters and biological states, the chicken coop plane is divided into multiple dynamic integrated control zones.

[0037] Furthermore, in the method provided in the application embodiment, by deploying thermal infrared cameras and visible light cameras covering the chicken coop to simultaneously collect infrared thermal imaging data streams and visible light video data streams, and fusing the global multimodal environmental dynamic thermal map for coupled cluster analysis, the chicken coop plane is divided into multiple dynamic comprehensive control zones, and the method further includes: A deep visual recognition model is used to fuse and analyze the infrared thermal imaging data stream and the visible light video data stream, outputting a heat map of chicken flock surface temperature distribution and a chicken flock behavior state vector; the global multimodal environmental dynamic heat map, the chicken flock surface temperature distribution heat map, and the chicken flock behavior state vector are fused to generate an environmental-biological coupling state matrix; by performing spatial clustering analysis on the environmental-biological coupling state matrix, the chicken house plane is divided into multiple dynamic integrated control zones.

[0038] In this embodiment, a deep vision recognition model is used to fuse and analyze infrared thermal imaging data streams and visible light video data streams. First, based on the thermal radiation characteristics of chickens, multiple frames of infrared thermal imaging in the data stream are segmented into independent chicken regions. Temperature extreme values ​​are extracted from the segmented chicken regions to form multi-frame chicken temperature distribution data. Then, Gaussian kernel density reconstruction is performed on the multi-frame chicken temperature distribution data to obtain a chicken surface temperature distribution heatmap characterizing the spatial distribution features of chicken body temperature. Simultaneously, a three-dimensional convolutional neural network is used to extract temporal spatial features from the visible light video data stream, generating a behavioral feature tensor containing chicken density distribution, motion vector field, and key posture features. Based on this, a cross-modal attention mechanism is used to fuse the behavioral feature tensor with the chicken surface temperature distribution heatmap, comprehensively characterizing the physiological and behavioral features of the chicken flock, ultimately generating a chicken flock behavioral state vector containing stress state, activity index, distribution evenness, and the proportion of abnormal behaviors.

[0039] Next, the global multimodal environmental dynamic heatmap, the chicken flock surface temperature distribution heatmap, and the chicken flock behavior state vector are integrated. In this process, a unified spatial coordinate network is first established based on the three-dimensional physical structure of the chicken house, and the global multimodal environmental dynamic heatmap is resampled onto this spatial coordinate network to obtain the environmental parameter representations for the corresponding spatial locations. Simultaneously, the chicken flock surface temperature distribution heatmap is registered to the spatial coordinate network using thin-plate spline interpolation, and chicken flock surface temperature information is extracted at the corresponding spatial locations. Then, the chicken flock behavior state vector is mapped according to the spatial coordinate locations of the chickens, and a continuous behavioral feature field is formed through Gaussian kernel diffusion, thereby obtaining the corresponding behavioral feature representations in the spatial coordinate network. Subsequently, the environmental parameters, chicken flock surface temperature, and behavioral features are combined under a unified spatial coordinate system to construct an environmental-biological coupling state tensor. Tensor dimensionality reduction is then performed on the environmental-biological coupling state tensor to finally form an environmental-biological coupling state matrix describing the coupling relationship between the chicken house environment and the chicken flock's biological state.

[0040] Finally, spatial clustering analysis was performed on the environmental-biological coupling state matrix to divide the chicken house plane into the multiple dynamic integrated control zones. In this process, firstly, using a spatial coordinate network as an index, the corresponding data in the environmental-biological coupling state matrix for each spatial location on the chicken house plane was read. This corresponding data is a combined expression of the environmental parameter characteristics, chicken body surface temperature characteristics, and chicken behavior characteristics at that spatial location at the same time. Then, a uniform scaling process was performed on each feature dimension in the environmental-biological coupling state matrix, calculating the minimum and maximum values ​​of each feature dimension across the entire chicken house. Next, the value of each location of the feature dimension was normalized by subtracting the minimum value from the current location value and dividing by the difference between the maximum and minimum values, allowing for direct comparison of the numerical ranges of different feature dimensions. Subsequently, several spatial locations on the chicken house plane were selected as initial representative locations for each zone, and the corresponding data of these initial representative locations in the environmental-biological coupling state matrix were used as the initial representative data for each zone. Next, for each spatial location within the chicken coop plane, the difference between its corresponding data and the representative data of each initial partition is calculated. The difference is calculated as the square root of the sum of the squared differences of each feature dimension, thus obtaining the set of difference values ​​for that spatial location relative to each partition. After obtaining the set of difference values, the spatial location is assigned to the partition with the smallest difference value. Combined with the adjacency constraints of the spatial coordinate network, the same partition is kept continuously distributed on the chicken coop plane. If an isolated spatial location is not adjacent to a region in the same partition, it is assigned to the partition with the second smallest difference value and connected to the adjacent spatial location. After completing one full spatial location division, the corresponding data of all spatial locations in each partition are averaged along each feature dimension in the environmental-biological coupling state matrix to obtain updated partition representative data. The difference value calculation and spatial location division steps are then repeated using the updated partition representative data until the results of two consecutive divisions no longer change or the change in the partition representative data is less than a preset threshold. Finally, the chicken house plane area corresponding to each stable zone is determined as a dynamic integrated control zone. This allows the chicken house plane to be divided into multiple dynamic integrated control zones by performing spatial clustering analysis on the environmental-biological coupling state matrix. The above calculation process is repeated when the environmental-biological coupling state matrix is ​​updated over time so that the dynamic integrated control zones can be dynamically adjusted according to changes in the environment and flock status.

[0041] Furthermore, the method provided in the application embodiment employs a deep vision recognition model to perform fusion analysis of the infrared thermal imaging data stream and the visible light video data stream, outputting a heat map of chicken flock surface temperature distribution and a chicken flock behavior state vector, and further includes: Based on the thermal radiation characteristics of chickens, independent chicken region segmentation and extreme temperature values ​​extraction are performed on multiple frames of infrared thermal imaging data stream to obtain multi-frame chicken temperature distribution data. Gaussian kernel density reconstruction is performed on the multi-frame chicken temperature distribution data to output a chicken surface temperature distribution heatmap. A three-dimensional convolutional neural network is driven to extract temporal spatial features from the visible light video data stream to obtain a behavioral feature tensor, which includes chicken density distribution, motion vector field, and key pose features. The behavioral feature tensor and the chicken surface temperature distribution heatmap are fused through a cross-modal attention mechanism to generate a chicken flock behavior state vector, which includes stress state, activity index, distribution uniformity, and abnormal behavior proportion.

[0042] In this embodiment, a deep vision recognition model is used to perform fusion analysis on infrared thermal imaging data streams and visible light video data streams to output a chicken flock surface temperature distribution heatmap and a chicken flock behavior state vector. Its structure includes an infrared thermal imaging feature extraction part for processing the infrared thermal imaging data stream, a visible light video feature extraction part for processing the visible light video data stream, and a cross-modal attention mechanism part for fusing the two types of features. The infrared thermal imaging feature extraction part is responsible for obtaining multi-frame chicken temperature distribution data from the infrared thermal imaging data stream and forming a chicken flock surface temperature distribution heatmap. The visible light video feature extraction part obtains a behavior feature tensor from the visible light video data stream through a three-dimensional convolutional neural network. The cross-modal attention mechanism part fuses the behavior feature tensor with the chicken flock surface temperature distribution heatmap to generate a chicken flock behavior state vector.

[0043] When performing independent chicken region segmentation and extracting extreme temperature values ​​from multiple frames of infrared thermal imaging data streams based on the thermal radiation characteristics of chickens, the process begins by reading multiple frames of infrared thermal imaging sequentially and completing temperature calibration, assigning a temperature value to each pixel. Then, based on the characteristic that the temperature of the chicken region is generally higher than that of the background region, the average temperature of the background region is calculated by statistically analyzing the pixel temperatures in the same frame of infrared thermal imaging. A preset temperature offset is then added to this average temperature as a segmentation threshold. All pixel temperature values ​​in the frame are then thresholded, with pixels exceeding the threshold marked as target pixels and the rest as background pixels. After obtaining the target pixels, connected regions are extracted from adjacent target pixels, merging them into several regions. Obvious noise regions are then removed using area and shape criteria, retaining regions that conform to the chicken's size characteristics as the independent chicken region segmentation result. In the process of extracting extreme temperature values ​​in chicken areas, for each independent chicken area, the temperature value of any pixel in that area is first read as the current highest and lowest temperature values. Then, the temperature values ​​of the remaining pixels in that area are read pixel by pixel and compared with the current highest temperature value. If the highest temperature value is higher, the current highest temperature value is updated. At the same time, it is compared with the current lowest temperature value. If the lowest temperature value is lower, the current lowest temperature value is updated. This process is repeated until all traversals are completed, thereby obtaining the highest and lowest temperature values ​​of that independent chicken area as the extreme temperature values ​​of the chicken area. The location of the independent chicken area is associated with the extreme temperature values ​​of the chicken area and recorded. The data is accumulated in a frame sequence to form multi-frame chicken temperature distribution data.

[0044] Next, when performing Gaussian kernel density reconstruction on the multi-frame chicken temperature distribution data, each independent chicken region is first converted into a temperature sample point. The location of the temperature sample point is represented by the geometric center of that independent chicken region in the image, and its representative temperature value is represented by the average of the highest and lowest temperature values ​​in that region. Then, a regular grid is established within the chicken coop plane, dividing the chicken coop plane into several equally spaced small regions. For each small region in the grid, the actual distance from the center of that small region to all temperature sample points is calculated sequentially, and the influence range is determined according to a fixed distance segmentation method. Temperature sample points within the first distance interval are included in the temperature estimation in full; those within the second distance interval are included in half; those within the third distance interval are included in one-quarter; and temperature sample points exceeding a preset maximum distance threshold are not included in the temperature estimation of that small region. The first, second, and third distance intervals are pre-defined. After determining the distance segmentation, the representative temperature values ​​of all temperature sample points involved in the calculation of that small area are accumulated according to their corresponding inclusion ratios, and then divided by the total proportion of the calculations to obtain the temperature estimate for that small area. By repeating the above distance determination, proportion inclusion, and averaging process for all small areas within the chicken house plane, a continuous temperature distribution result covering the entire chicken house plane is finally obtained. This continuous temperature distribution result is then output in the form of a two-dimensional heat map, forming a surface temperature distribution heat map of the chicken flock.

[0045] Subsequently, when driving the 3D convolutional neural network to extract temporal spatial features from the visible light video data stream, continuous video frames are first extracted from the visible light video data stream at fixed time intervals to form a frame sequence containing temporal continuity. Resolution unification and brightness normalization are then performed on the frame sequence. The frame sequence is then input into the 3D convolutional neural network, which extracts spatial features such as the chicken flock's appearance outline and positional distribution from each frame in the spatial direction. Simultaneously, it analyzes the changes between adjacent frames in the temporal direction. This analysis process involves treating several consecutive frames as a processing segment. Within each frame, a fixed-size sliding window scans the image block by block, calculating local texture and edge changes for each region to form its spatial features. The spatial features of all regions in the entire frame are then summarized to form the spatial feature representation of that frame. Then, at the same spatial location, the spatial feature representations of adjacent frames are differentially analyzed position by position to obtain the change amount. The changes are then accumulated within the processing segment to obtain the temporal change intensity, ensuring that each spatial location simultaneously possesses both spatial features and temporal change intensity representations. Based on this, the chicken house plane is divided into several equal-area regions, and the number of target responses is accumulated in the corresponding regions to form the chicken density distribution. The positive and negative directions of the change in the same spatial position in adjacent time slices are taken as the direction of change, and the absolute value of the change is taken as the amplitude of change to form a motion vector field. The spatial positions with significant contour changes in consecutive frames and their frequency of occurrence and duration in the processed segments are summarized to form key posture features, thereby obtaining a behavioral feature tensor that includes chicken density distribution, motion vector field and key posture features.

[0046] Finally, when fusing the behavioral feature tensor and the chicken flock surface temperature distribution heatmap using a cross-modal attention mechanism, the behavioral feature tensor and the chicken flock surface temperature distribution heatmap are first spatially aligned to ensure a one-to-one correspondence between the two types of data at the same spatial location on the chicken house plane. Then, the behavioral feature values ​​and surface temperature values ​​corresponding to each spatial location are sequentially read on the chicken house plane, and the behavioral and temperature changes at that location between the current and previous times are calculated. The two types of changes are then uniformly converted to the same numerical range to ensure comparability. Based on this, the direction of change of behavioral and temperature changes is compared for each spatial location. If both are increasing or decreasing, the direction is considered consistent; if one increases while the other decreases, the direction is considered inconsistent. In cases of consistent direction, the behavioral change at that spatial location is directly used as the fusion input; in cases of inconsistent direction, the behavioral change at that spatial location is set to zero, thus preventing that spatial location from affecting the fusion results in subsequent statistics. After completing the orientation consistency processing, the spatial locations with consistent orientations are numerically adjusted based on the difference between the behavioral change and the temperature change. The maximum allowable difference is determined by the maximum difference between the behavioral change and the temperature change in historical operating data. When the difference is less than half of the maximum allowable difference, the original behavioral change is retained. When the difference is greater than half of the maximum allowable difference but less than the maximum allowable difference, the behavioral change is linearly reduced according to the proportion of the difference to the maximum allowable difference. When the difference is equal to or exceeds the maximum allowable difference, the behavioral change is set to zero. This ensures that the spatial locations with the closest behavioral change to the temperature change retain the most complete behavioral change in the fusion result. After completing the above processing of all spatial locations on the chicken house floor plan, a fused set of spatial location features is obtained. These features are then statistically summarized across the entire chicken house. The stress state is determined by comparing the number of spatial locations where both the behavior change and temperature change exceed the normal fluctuation threshold with the total number of spatial locations. The activity index is obtained by averaging the behavior changes of all spatial locations after fusion. The distribution uniformity is obtained by calculating and normalizing the deviation of the behavior changes of each spatial location from the overall average. The proportion of abnormal behavior is obtained by statistically analyzing the percentage of spatial locations where the behavior changes significantly deviate from the overall average. Finally, the stress state, activity index, distribution uniformity, and abnormal behavior proportion are combined to form a flock behavior state vector.

[0047] Furthermore, the method provided in the application embodiment, which integrates the global multimodal environmental dynamic heat map, the chicken flock surface temperature distribution heat map, and the chicken flock behavior state vector to generate an environmental-biological coupling state matrix, further includes: A spatial coordinate network for the chicken house is established based on the three-dimensional physical structure. The global multimodal environmental dynamic heat map is resampled to the spatial coordinate network to extract multiple environmental parameter vectors from multiple grid cells. The chicken flock surface temperature distribution heat map is registered to the spatial coordinate network using thin-plate spline interpolation to extract multiple chicken flock surface temperature values ​​from multiple grid cells. The chicken flock behavior state vector is mapped according to the chicken coordinate position. After generating a behavior feature field through Gaussian kernel diffusion, it is combined with the spatial coordinate network to decompose and obtain multiple behavior feature vectors from multiple grid cells. Multiple environmental parameter vectors, multiple chicken flock surface temperature values, and multiple behavior feature vectors are combined to construct multiple environmental-biological coupling state tensors. Tensor dimensionality reduction is performed on the multiple environmental-biological coupling state tensors to output the environmental-biological coupling state matrix.

[0048] In this embodiment of the application, when establishing the spatial coordinate network of the chicken house based on the three-dimensional physical structure, the three-dimensional physical structure information of the chicken house is first read to obtain the length, width, height of the chicken house, and the position boundaries of the walls, ceiling, ground and partitions. Then, the origin of the spatial coordinates is selected and the direction of the coordinate axes is determined so that any point inside the chicken house can be represented by spatial coordinates. Subsequently, a set of regular coordinate points is generated in the length and width directions according to the preset spatial spacing, and the set of regular coordinate points is matched with the boundary of the chicken house to finally form a spatial coordinate network covering the plane of the chicken house.

[0049] To extract multiple environmental parameter vectors from multiple grid cells by resampling the global multimodal environmental dynamic heatmap onto a spatial coordinate network, a resampling method is used to perform coordinate mapping on the global multimodal environmental dynamic heatmap. In this process, for each coordinate point in the spatial coordinate network, the corresponding spatial location in the global multimodal environmental dynamic heatmap is found, and the values ​​of temperature, humidity, ammonia, hydrogen sulfide, carbon dioxide, and dust concentration are read from that location. If the coordinate point falls among discrete points in the global multimodal environmental dynamic heatmap, neighborhood interpolation is used to calculate the value of that coordinate point from the values ​​of several surrounding points. Subsequently, the environmental parameter values ​​at the same coordinate point are combined in a fixed order to obtain the environmental parameter vector corresponding to that coordinate point, thus forming multiple environmental parameter vectors on multiple grid cells of the spatial coordinate network.

[0050] Subsequently, the chicken body surface temperature distribution heatmap was registered to the spatial coordinate network using thin-plate spline interpolation. When extracting multiple chicken body surface temperature values ​​from multiple grid cells, the thin-plate spline interpolation method was used to register the chicken body surface temperature distribution heatmap, establishing a correspondence between the temperature pixel coordinates in the chicken body surface temperature distribution heatmap and the coordinate points in the spatial coordinate network. The known temperature points in the chicken body surface temperature distribution heatmap were used as interpolation control points. A smooth temperature surface was generated in space through thin-plate spline interpolation, so that each coordinate point in the spatial coordinate network could obtain a corresponding temperature value on the temperature surface. Then, the temperature value was read point by point and recorded as the chicken body surface temperature value of that grid cell, thus forming multiple chicken body surface temperature values ​​on multiple grid cells.

[0051] Next, the chicken behavior state vectors are mapped according to the chicken's coordinate position. After generating a behavior feature field through Gaussian kernel diffusion, multiple behavior feature vectors are obtained by decomposing the spatial coordinate network into multiple grid cells. In this process, the Gaussian kernel diffusion method is used to spatially diffuse the chicken behavior state vectors. According to the chicken's coordinate position, the chicken behavior state vector is placed at the corresponding coordinate point in the spatial coordinate network, so that the spatial position of each chicken is associated with its corresponding chicken behavior state vector. Then, with each chicken's coordinate point as the center, the influence of its chicken behavior state vector is diffused to the surrounding coordinate points in the spatial coordinate network. The diffusion intensity decreases according to a Gaussian function as the distance between the coordinate point and the chicken's coordinate point increases. The diffusion results from multiple chickens at the same coordinate point are accumulated to obtain the behavior feature field value of that coordinate point. After completing the diffusion accumulation of the entire spatial coordinate network, the behavior feature field value at the corresponding coordinate point of each grid cell is read and combined in a fixed order to form the behavior feature vector of that grid cell, thus obtaining multiple behavior feature vectors on multiple grid cells.

[0052] Subsequently, when combining multiple environmental parameter vectors, multiple chicken population surface temperature values, and multiple behavioral feature vectors, a feature concatenation method is used to perform the combination on each grid cell of the spatial coordinate network. For the same grid cell, the environmental parameter vector, chicken population surface temperature value, and behavioral feature vector are read simultaneously, and the chicken population surface temperature value is inserted between the environmental parameter vector and the behavioral feature vector at a preset position, so that the three types of information form a unified multidimensional data structure within the same grid cell. Then, the above combination process is repeated for all grid cells in the spatial coordinate network, thereby constructing multiple environmental-biological coupled state tensors indexed by grid cells.

[0053] Finally, when performing tensor dimensionality reduction on multiple environmental-biological coupled state tensors, the following steps are taken: First, for each grid cell in the spatial coordinate network, its corresponding environmental-biological coupled state tensor is read, and the environmental parameter values, chicken population surface temperature values, and behavioral characteristic values ​​are arranged in a fixed order to form a numerical list, ensuring that each grid cell corresponds to a numerical list of the same length. Then, the numerical lists of all grid cells are summarized to form an overall data table, and the average value of each column in the data table is calculated for all grid cells. This average value is then subtracted from the value of each column in each grid cell, ensuring that all data are compared around the same benchmark. After this processing, the spatial variation of the values ​​in each column of the data table is sorted, and the columns with the largest variations are selected as the primary features; the columns with smaller variations are not retained separately. Next, for each grid cell, only the values ​​in its primary feature columns are retained, and a new short list is formed in the original order, simplifying the environmental-biological coupled state tensor, which originally contained multiple data points, into a representation containing only a small number of key data points. Finally, according to the arrangement order of the grid cells in the spatial coordinate network, the simplified data lists of all grid cells are arranged sequentially to form the environmental-biological coupled state matrix.

[0054] Step S400: Based on the multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, set multiple environmental regulation optimization targets, perform rolling optimization of environmental parameters, and output multiple time-series control parameter groups.

[0055] In this embodiment, based on the environmental deviation vector and biological stress vector formed within multiple dynamic integrated control zones, the environmental control needs of different zones are quantitatively analyzed to determine the environmental regulation priority corresponding to each zone, and an environmental regulation optimization target matching the zone status is set. Subsequently, based on the environmental deviation vector and biological stress vector, multiple sets of selectable environmental regulation parameters are matched in the environmental regulation parameter space. Finally, using the environmental regulation optimization target as a constraint, rolling optimization calculations are performed on multiple sets of environmental regulation parameters, so that the environmental regulation parameters are continuously updated over time, and multiple sets of time-series control parameters are output to guide the chicken house environmental control equipment to execute in a time sequence.

[0056] Furthermore, the method provided in the application embodiment, which sets multiple environmental regulation and optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated regulation zones, performs rolling optimization of environmental parameter tuning, and outputs multiple time-series control parameter groups, further includes: Based on the multiple environmental deviation vectors and multiple biological stress vectors, multiple environmental regulation priorities are quantified; multiple environmental regulation optimization objectives are set according to the multiple environmental regulation priorities; multiple sets of environmental regulation parameters are matched according to the multiple environmental deviation vectors and multiple biological stress vectors; the multiple sets of environmental regulation parameters are solved by rolling optimization according to the multiple environmental regulation optimization objectives, and the multiple sets of time-series control parameters are output.

[0057] In this embodiment, when quantifying the priorities of multiple environmental adjustments, for each dynamic integrated control zone, firstly, the environmental deviation vector corresponding to that zone is obtained, and temperature deviation values ​​and humidity deviation values ​​are read from it. The temperature deviation values ​​and humidity deviation values ​​are the numerical differences between the current monitored values ​​and the corresponding target control interval boundaries. Simultaneously, the biological stress vector corresponding to that zone is obtained, and multiple chicken flock surface temperature values ​​are read from it. The average of the multiple chicken flock surface temperature values ​​is then calculated to obtain the average surface temperature. This average surface temperature is then compared with a pre-set normal chicken flock surface temperature control interval to obtain the surface temperature deviation value. Further, the corresponding values ​​from multiple behavioral feature vectors are read... Stress status, activity index, distribution evenness, and proportion of abnormal behavior are measured and their differences are calculated with corresponding normal reference thresholds to obtain behavioral deviation values. The body surface temperature deviation value is added to the behavioral deviation value to obtain the biological stress quantity. Subsequently, the product of the temperature deviation value and the biological stress quantity, as well as the product of the humidity deviation value and the biological stress quantity, are calculated separately. The two product results are compared. When the product value corresponding to temperature is greater than the product value corresponding to humidity, the environmental regulation priority for that zone is determined to be temperature regulation first. When the product value corresponding to humidity is greater than the product value corresponding to temperature, the environmental regulation priority for that zone is determined to be humidity regulation first. This completes the quantification of multiple environmental regulation priorities.

[0058] Next, when setting multiple environmental regulation optimization objectives based on multiple environmental regulation priorities, for each dynamic integrated control zone, a corresponding set of environmental regulation optimization objectives is set based on the determined environmental regulation priorities of that zone. Specifically, when the environmental regulation priority of a dynamic integrated control zone is to prioritize temperature regulation, the first environmental regulation optimization objective for that zone is to reduce the temperature deviation to zero and maintain it within the target control range, while the second environmental regulation optimization objective is to control the humidity deviation within a preset allowable range. When the environmental regulation priority of a dynamic integrated control zone is to prioritize humidity regulation, the first environmental regulation optimization objective for that zone is to reduce the humidity deviation to zero and maintain it within the target control range, while the second environmental regulation optimization objective is to control the temperature deviation within a preset allowable range. In addition, in each dynamic integrated control zone, the biological stress level is set to not increase during the regulation process as a constraint-type environmental regulation optimization objective, thereby obtaining multiple environmental regulation optimization objectives covering multiple dynamic integrated control zones.

[0059] Subsequently, when matching multiple sets of environmental regulation parameters based on multiple environmental deviation vectors and multiple biological stress vectors, for each dynamic integrated regulation zone, the regulation directions for temperature and humidity are first determined according to the environmental deviation vectors. When the corresponding environmental parameter is higher than the upper limit of the target control interval, the regulation direction is determined to be decreasing; when it is lower than the lower limit of the target control interval, the regulation direction is determined to be increasing. Then, the widths of the target control intervals for temperature and humidity are calculated separately, and fixed-proportion regulation amplitudes are generated based on their respective interval widths. The regulation amplitudes are taken as one-quarter, one-half, and three-quarters of the target control interval width, respectively. For environmental parameters corresponding to environmental regulation priorities, multiple regulation values ​​are generated using the above three regulation amplitudes. For environmental parameters that are not prioritized for regulation, only the regulation amplitude corresponding to one-quarter of the interval width is used to generate regulation values. Then, the different regulation values ​​are combined to form multiple sets of environmental regulation parameters, ensuring that each set of environmental regulation parameters includes clearly defined temperature and humidity regulation values.

[0060] Finally, when performing rolling optimization of multiple sets of environmental regulation parameters based on multiple environmental regulation optimization objectives, for each dynamic integrated control zone, within the current control cycle, the effects of each of the matched sets of environmental regulation parameters are calculated one by one, and the set that best meets the multiple environmental regulation optimization objectives is selected as the execution result. Specifically, firstly, the actual environmental parameter values ​​of the dynamic integrated control zone in the current control cycle are read, including the current temperature and humidity values, and the corresponding biological stress of the zone in the current control cycle is read as the baseline value. Then, for each set of environmental regulation parameters, the current actual environmental parameter value is directly added or subtracted according to the given temperature and humidity adjustment values ​​in the set of parameters. When the adjustment direction is decreasing, the adjustment value is subtracted from the current value; when the adjustment direction is increasing, the adjustment value is added to the current value, thereby obtaining the adjusted temperature and humidity values ​​corresponding to the set of environmental regulation parameters.

[0061] After obtaining the adjusted temperature and humidity values, the adjusted environmental deviation values ​​corresponding to this set of environmental regulation parameters are recalculated. This involves comparing the adjusted temperature and humidity values ​​with their respective target control intervals to calculate new temperature and humidity deviation values. Simultaneously, biological stress is updated based on changes in environmental deviation. When the adjusted temperature and humidity deviation values ​​decrease compared to the current control period, it is assumed that the body surface temperature deviation and behavioral deviation values ​​have decreased synchronously, and the biological stress is reduced accordingly in the same direction. When the adjusted environmental deviation values ​​do not decrease or increase, the biological stress is considered to remain unchanged or increase, thus obtaining the adjusted biological stress corresponding to this set of environmental regulation parameters. After completing the above calculations for each set of environmental regulation parameters, the adjusted temperature deviation, adjusted humidity deviation, and adjusted biological stress are compared with the multiple environmental regulation optimization targets one by one. The set of environmental regulation parameters that satisfies the following conditions—a decrease in the environmental deviation corresponding to the first environmental regulation optimization target, the environmental deviation corresponding to the second environmental regulation optimization target not exceeding the allowable range, and the adjusted biological stress not exceeding the current biological stress—is selected as the execution parameter for the current control cycle. This execution parameter is then recorded as the parameter value of the time-series control parameter set for the current control cycle, and the process proceeds to the next control cycle. In the next control cycle, the updated environmental deviation vector and biological stress vector are re-acquired, and the processes of environmental regulation priority quantification, environmental regulation optimization target setting, matching of multiple sets of environmental regulation parameters, and calculation and selection of regulation effects are repeated. This ensures that the environmental regulation parameters are continuously updated with each control cycle, ultimately forming multiple time-series control parameter sets composed of execution parameters from multiple consecutive control cycles.

[0062] Step S500: Based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, perform conflict detection and fusion of the multiple time-series control parameter groups, output a linkage control strategy, and execute intelligent control of the chicken house's breeding environment.

[0063] In this embodiment, when performing conflict detection and fusion on multiple time-series control parameter groups based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, the layout information of all environmental control equipment in the chicken house is first read to clarify the installation location, controllable parameter type, and actual range of action of each environmental control device, thereby determining the specific spatial area that each environmental control device can affect. Simultaneously, the spatial locations of multiple dynamic integrated control zones in the chicken house are read, and the spatial association topology between each dynamic integrated control zone is determined based on whether the zones are adjacent and whether there is mutual environmental influence.

[0064] Subsequently, multiple time-series control parameter groups are read in chronological order. At each time node, the control parameters corresponding to each dynamic integrated control zone are assigned to the environmental control equipment that can cover that zone, so that each environmental control equipment forms a corresponding set of control parameters at that time node.

[0065] Next, conflict detection is performed on the set of control parameters within the same time node to determine whether the same environmental control device simultaneously receives control parameters from multiple different zones, and whether these control parameters are opposite in adjustment direction or significantly inconsistent in adjustment magnitude, or whether adjacent dynamic integrated control zones produce mutually canceling adjustment effects on the same spatial area through different devices. When a conflict is detected, the control parameters corresponding to the dynamic integrated control zone with higher environmental adjustment priority are prioritized according to the spatial correlation topology, and the remaining conflicting control parameters are merged or reduced to ensure that the final retained control parameters are consistent in adjustment direction and do not produce spatial offsetting. After completing the conflict detection and parameter fusion for each time node, the obtained fused control parameters are organized in chronological order to form a unified linkage control strategy. Finally, the linkage control strategy is sent to each environmental control device, which then coordinates and executes the corresponding adjustment operations in a unified chronological order, thereby achieving continuous, coordinated, and intelligent control of the breeding environment throughout the entire chicken house.

[0066] In summary, the embodiments of this application have at least the following technical effects: This application deploys a distributed multimodal environmental sensing array in a chicken house based on a preset radius scale; it receives microbial concentration reference information returned from the laboratory as a baseline truth value, aligns it with the multimodal environmental sensing data returned from the multimodal environmental sensing array, performs spatial interpolation prediction based on fluid dynamics simulation, and constructs a global multimodal environmental dynamic heat map; it deploys thermal infrared cameras and visible light cameras covering the chicken house to simultaneously collect infrared thermal imaging data streams and visible light video data streams, fuses the global multimodal environmental dynamic heat map for coupled cluster analysis, and divides the chicken house plane into multiple dynamic integrated control zones; it sets multiple environmental regulation optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, performs rolling optimization of environmental parameters, and outputs multiple time-series control parameter groups; based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, it performs conflict detection and fusion of the multiple time-series control parameter groups, outputs a linkage control strategy, and executes intelligent control of the chicken house's breeding environment. This invention addresses the technical problems in existing technologies, such as the inability to accurately perceive the spatial distribution of the breeding environment, the coarse control strategies, and the difficulty in dynamically matching the actual needs of chicken flocks. By constructing a global multimodal dynamic heat map of the environment and performing dynamic partitioning and rolling optimization control accordingly, the invention achieves the technical effect of refined, dynamic, and intelligent control of the breeding environment.

[0067] Example 2, based on the same inventive concept as the IoT-based intelligent control method for aquaculture environments described in the preceding examples, such as... Figure 2 As shown, this application provides an intelligent control system for aquaculture environments based on the Internet of Things (IoT). The system and method embodiments in this application are based on the same inventive concept. The system includes: Deployment module 11 is used for distributed deployment of a multimodal environment sensing array in the chicken house based on a preset radius scale; prediction module 12 is used to receive microbial concentration reference information returned from the laboratory as a baseline truth value, align with the multimodal environment sensing data returned by the multimodal environment sensing array, perform spatial interpolation prediction based on fluid dynamics simulation, and construct a global multimodal environment dynamic heat map; clustering analysis module 13 is used to simultaneously collect infrared thermal imaging data streams and visible light video data streams by deploying thermal infrared cameras and visible light cameras covering the chicken house, and fuse the global multimodal environment... The dynamic heatmap is used for coupled cluster analysis to divide the chicken house into multiple dynamic integrated control zones. The rolling optimization module 14 is used to set multiple environmental regulation optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, perform rolling optimization of environmental parameters, and output multiple time-series control parameter groups. The intelligent control module 15 is used to perform conflict detection and fusion of the multiple time-series control parameter groups based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, output linkage control strategies, and execute intelligent control of the breeding environment of the chicken house.

[0068] Furthermore, the system is also used to implement the following functions: A deep visual recognition model is used to fuse and analyze the infrared thermal imaging data stream and the visible light video data stream, outputting a heat map of chicken flock surface temperature distribution and a chicken flock behavior state vector; the global multimodal environmental dynamic heat map, the chicken flock surface temperature distribution heat map, and the chicken flock behavior state vector are fused to generate an environmental-biological coupling state matrix; by performing spatial clustering analysis on the environmental-biological coupling state matrix, the chicken house plane is divided into multiple dynamic integrated control zones.

[0069] Furthermore, the system is also used to implement the following functions: Based on the thermal radiation characteristics of chickens, independent chicken region segmentation and extreme temperature values ​​extraction are performed on multiple frames of infrared thermal imaging data stream to obtain multi-frame chicken temperature distribution data. Gaussian kernel density reconstruction is performed on the multi-frame chicken temperature distribution data to output a chicken surface temperature distribution heatmap. A three-dimensional convolutional neural network is driven to extract temporal spatial features from the visible light video data stream to obtain a behavioral feature tensor, which includes chicken density distribution, motion vector field, and key pose features. The behavioral feature tensor and the chicken surface temperature distribution heatmap are fused through a cross-modal attention mechanism to generate a chicken flock behavior state vector, which includes stress state, activity index, distribution uniformity, and abnormal behavior proportion.

[0070] Furthermore, the system is also used to implement the following functions: Each multimodal environmental sensing node in the multimodal environmental sensing array supports real-time sensing of temperature, humidity, ammonia, hydrogen sulfide, carbon dioxide, and dust concentration. The global multimodal environmental dynamic heat map includes a temperature field, a humidity field, a harmful gas concentration field, a dust concentration field, and a microbial risk prediction field.

[0071] Furthermore, the system is also used to implement the following functions: After spatiotemporal alignment of the multimodal environmental sensing data based on the multimodal environmental sensing array, the transient fluid field is solved by inputting it into the aquaculture environment fluid dynamics simulation model, and the temperature field, humidity field, harmful gas concentration field, and dust concentration field are output. Based on the statistical correlation between microbial concentration and environmental parameters in the chicken house, a risk probability model is constructed. Combining the temperature field, humidity field, harmful gas concentration field, and dust concentration field, the risk probability model is used to extrapolate the microbial risk probability of the microbial concentration reference information, and the microbial risk prediction field is generated.

[0072] Furthermore, the system is also used to implement the following functions: Based on the three-dimensional physical structure of the chicken house, the layout of the ventilation equipment, and the performance information of the ventilation equipment, a parametric fluid dynamics mesh model is established. Using the spatiotemporally aligned microbial concentration reference information and multimodal environmental perception data as joint constraints, the dynamic boundary parameters of the parametric fluid dynamics mesh model are calibrated in reverse through a data assimilation algorithm, and a simulation model of the aquaculture environment fluid dynamics is output.

[0073] Furthermore, the system is also used to implement the following functions: A spatial coordinate network for the chicken house is established based on the three-dimensional physical structure. The global multimodal environmental dynamic heat map is resampled to the spatial coordinate network to extract multiple environmental parameter vectors from multiple grid cells. The chicken flock surface temperature distribution heat map is registered to the spatial coordinate network using thin-plate spline interpolation to extract multiple chicken flock surface temperature values ​​from multiple grid cells. The chicken flock behavior state vector is mapped according to the chicken coordinate position. After generating a behavior feature field through Gaussian kernel diffusion, it is combined with the spatial coordinate network to decompose and obtain multiple behavior feature vectors from multiple grid cells. Multiple environmental parameter vectors, multiple chicken flock surface temperature values, and multiple behavior feature vectors are combined to construct multiple environmental-biological coupling state tensors. Tensor dimensionality reduction is performed on the multiple environmental-biological coupling state tensors to output the environmental-biological coupling state matrix.

[0074] Furthermore, the system is also used to implement the following functions: Based on the multiple environmental deviation vectors and multiple biological stress vectors, multiple environmental regulation priorities are quantified; multiple environmental regulation optimization objectives are set according to the multiple environmental regulation priorities; multiple sets of environmental regulation parameters are matched according to the multiple environmental deviation vectors and multiple biological stress vectors; the multiple sets of environmental regulation parameters are solved by rolling optimization according to the multiple environmental regulation optimization objectives, and the multiple sets of time-series control parameters are output.

[0075] Example 3: Based on the inventive concept of the IoT-based intelligent control method for aquaculture environment in the foregoing examples, this application also provides an electronic device, including: at least one processor; a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of any of the methods described in Example 1 above.

[0076] Figure 3 This is a schematic diagram of the structure of an exemplary electronic device of this application. Figure 3In this document, the bus architecture is represented by bus 300. Bus 300 may include any number of interconnected buses and bridges, and bus 300 connects various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, i.e., a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 302 is responsible for managing bus 300 and general processing, while memory 304 can be used to store data used by processor 302 during operation.

[0077] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent control of aquaculture environment based on the Internet of Things, characterized in that, The method includes: A distributed deployment of multimodal environmental sensing arrays is carried out in the chicken house based on a preset radius scale; The reference information on microbial concentration returned from the laboratory is used as the baseline truth value. The multimodal environmental sensing data returned from the multimodal environmental sensing array is aligned with the data. Spatial interpolation prediction based on fluid dynamics simulation is performed to construct a global multimodal environmental dynamic heat map. By deploying thermal infrared cameras and visible light cameras covering the chicken house to simultaneously collect infrared thermal imaging data streams and visible light video data streams, and fusing the global multimodal environment dynamic heat map for coupled cluster analysis, the chicken house plane is divided into multiple dynamic comprehensive control zones. Based on the multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated control zones, multiple environmental regulation and optimization targets are set, environmental parameter tuning and rolling optimization are performed, and multiple time-series control parameter groups are output. Based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, conflict detection and fusion of the multiple time-series control parameter groups are performed, and a linkage control strategy is output to execute intelligent control of the chicken house's breeding environment.

2. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 1, characterized in that, By deploying thermal infrared cameras and visible light cameras covering the chicken coop to simultaneously collect infrared thermal imaging data streams and visible light video data streams, and fusing the global multimodal environmental dynamic thermal map for coupled cluster analysis, the chicken coop plane is divided into multiple dynamic integrated control zones. The method includes: A deep visual recognition model is used to perform fusion analysis of the infrared thermal imaging data stream and the visible light video data stream, and outputs a heat map of the chicken flock's surface temperature distribution and a chicken flock behavior state vector. By integrating the global multimodal environmental dynamic heat map, the chicken flock surface temperature distribution heat map, and the chicken flock behavior state vector, an environmental-biological coupling state matrix is ​​generated. By performing spatial cluster analysis on the environmental-biological coupling state matrix, the chicken house plane is divided into multiple dynamic integrated control zones.

3. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 2, characterized in that, A deep visual recognition model is used to fuse and analyze the infrared thermal imaging data stream and the visible light video data stream, outputting a heat map of chicken flock surface temperature distribution and a chicken flock behavior state vector. The method includes: Based on the thermal radiation characteristics of chickens, the multi-frame infrared thermal imaging data stream is used to perform independent chicken region segmentation and chicken region temperature extreme value extraction to obtain multi-frame chicken temperature distribution data. Gaussian kernel density reconstruction is performed on the multi-frame chicken temperature distribution data to output a heat map of chicken body surface temperature distribution; A three-dimensional convolutional neural network is driven to extract the temporal and spatial features of the visible light video data stream, resulting in a behavioral feature tensor, wherein the behavioral feature tensor includes chicken density distribution, motion vector field, and key pose features; The behavioral feature tensor and the chicken flock surface temperature distribution heatmap are fused by a cross-modal attention mechanism to generate the chicken flock behavioral state vector, wherein the chicken flock behavioral state vector includes stress state, activity index, distribution uniformity and abnormal behavior proportion.

4. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 1, characterized in that, Each multimodal environmental sensing node in the multimodal environmental sensing array supports real-time sensing of temperature, humidity, ammonia, hydrogen sulfide, carbon dioxide, and dust concentration. The global multimodal environmental dynamic heat map includes a temperature field, a humidity field, a harmful gas concentration field, a dust concentration field, and a microbial risk prediction field.

5. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 4, characterized in that, The method involves receiving microbial concentration reference information from the laboratory as a baseline, aligning it with the multimodal environmental sensing data returned by the multimodal environmental sensing array, performing spatial interpolation prediction based on fluid dynamics simulation, and constructing a global multimodal environmental dynamic heat map. The method includes: After performing spatiotemporal alignment of the multimodal environmental sensing data based on the multimodal environmental sensing array, the data is input into the aquaculture environment fluid dynamics simulation model to solve the transient fluid field, and the temperature field, humidity field, harmful gas concentration field, and dust concentration field are output. A risk probability model was constructed based on the statistical correlation between microbial concentration and environmental parameters in the chicken house. By combining the temperature field, humidity field, harmful gas concentration field, and dust concentration field, the risk probability model is used to extrapolate the microbial risk probability of the microbial concentration reference information, thereby generating the microbial risk prediction field.

6. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 5, characterized in that, The method further includes: Based on the three-dimensional physical structure of the chicken house, the layout of the ventilation equipment, and the performance information of the ventilation equipment, a parametric fluid dynamics mesh model was established. Using the spatiotemporally aligned microbial concentration reference information and multimodal environmental perception data as joint constraints, the dynamic boundary parameters of the parameterized fluid dynamics grid model are calibrated in reverse through a data assimilation algorithm, and a fluid dynamics simulation model of the aquaculture environment is output.

7. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 6, characterized in that, By integrating the global multimodal environmental dynamic heatmap, the chicken flock surface temperature distribution heatmap, and the chicken flock behavior state vector, an environmental-biological coupled state matrix is ​​generated. The method includes: Establish a spatial coordinate network for the chicken coop based on the described three-dimensional physical structure; The global multimodal environment dynamic heatmap is resampled to the spatial coordinate network to extract multiple environmental parameter vectors from multiple grid cells; The surface temperature distribution heatmap of the chicken flock is registered to the spatial coordinate network by thin plate spline interpolation, and multiple surface temperature values ​​of the chicken flock are extracted from the multiple grid cells; The chicken flock behavior state vector is mapped according to the chicken coordinate position. After generating the behavior feature field through Gaussian kernel diffusion, it is combined with the spatial coordinate network decomposition to obtain multiple behavior feature vectors of the multiple grid cells. By combining the multiple environmental parameter vectors, multiple chicken population surface temperature values, and multiple behavioral feature vectors, multiple environmental-biological coupled state tensors are constructed. Tensor dimensionality reduction is performed on the multiple environmental-biological coupling state tensors to output the environmental-biological coupling state matrix.

8. The intelligent control method for aquaculture environment based on the Internet of Things as described in claim 7, characterized in that, Based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated regulation zones, multiple environmental regulation and optimization objectives are set, environmental parameter tuning and rolling optimization are performed, and multiple time-series control parameter groups are output. The method includes: Based on the multiple environmental deviation vectors and multiple biological stress vectors, multiple environmental regulation priorities are quantified; The multiple environmental regulation optimization objectives are set based on the multiple environmental regulation priorities; Multiple sets of environmental regulation parameters are matched based on the multiple environmental deviation vectors and multiple biological stress vectors; Based on the multiple environmental regulation optimization objectives, the multiple sets of environmental regulation parameters are solved by rolling optimization, and the multiple sets of timing control parameters are output.

9. An intelligent control system for aquaculture environment based on the Internet of Things, characterized in that, The system is used to execute the IoT-based intelligent control method for aquaculture environments as described in any one of claims 1-8, and the system includes: The deployment module is used for the distributed deployment of multimodal environmental sensor arrays in the chicken house based on a preset radius scale; The prediction module is used to receive the microbial concentration reference information returned from the laboratory as the baseline truth value, align it with the multimodal environmental sensing data returned from the multimodal environmental sensing array, perform spatial interpolation prediction based on fluid dynamics simulation, and construct a global multimodal environmental dynamic heat map. The clustering analysis module is used to simultaneously collect infrared thermal imaging data streams and visible light video data streams by deploying thermal infrared cameras and visible light cameras covering the chicken house, and to perform coupled clustering analysis by fusing the global multimodal environment dynamic heat map to divide the chicken house plane into multiple dynamic comprehensive control zones. The rolling optimization module is used to set multiple environmental regulation optimization targets based on multiple environmental deviation vectors and multiple biological stress vectors of the multiple dynamic integrated regulation zones, perform rolling optimization of environmental parameters, and output multiple time-series control parameter groups; The intelligent control module is used to perform conflict detection and fusion of multiple time-series control parameter groups based on the layout of environmental control equipment in the chicken house and the spatial association topology of the multiple dynamic integrated control zones, output a linkage control strategy, and execute intelligent control of the breeding environment in the chicken house.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the IoT-based intelligent control method for aquaculture environment as described in any one of claims 1-8.

Images