Pig house environment regulation and control optimization method and system

Abstract

The invention discloses a pigsty environment regulation and control optimization method and system, and belongs to the technical field of breeding environment management. The system comprises an intelligent sensing module, an environment analysis module, a parameter optimization module and a control management module. The intelligent sensing module is used for collecting pig house environment parameters and live pig physiological parameters through a multi-mode sensor; the environment analysis module is used for constructing a behavior recognition model, analyzing behavior characteristics based on physiological parameters and dividing abnormal regions; the parameter optimization module is used for establishing a physiological environment control model and setting an optimal optimization scheme for each abnormal region; and the control management module is used for generating an environment control instruction according to the optimal optimization scheme and controlling equipment execution. The pig house environment and the pig physiological parameters are collected and intelligently analyzed through the multi-mode sensor, automatic and precise environment optimization management is achieved, and the comfort level and the health level of the pigs are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of breeding environment management, in particular to a pig house environment regulation optimization method and system. BACKGROUND

[0002] With the development of large-scale pig breeding, the precise ventilation mode is more in line with the current production needs of pig farms than the traditional tunnel ventilation mode and its derived ventilation modes. The body surface wind speed of pigs can be increased to reduce the body temperature of pigs and relieve heat stress, thereby improving production performance.

[0003] At present, the precise ventilation system usually relies on manual adjustment or mechanical temperature threshold control. This method has certain drawbacks, for example, the ventilation parameter adjustment requires personnel to continuously observe the performance of the pig group and manually intervene, the response is lagging and the operation consistency is poor, which restricts the improvement of production efficiency. The traditional environmental control device takes temperature as the core single-factor control target and relies on the classical linear control method, which cannot adapt to the complexity of the nonlinear time-varying system of the pig house, resulting in mismatch between the ventilation parameters and the actual needs. There is a nonlinear coupling relationship between the environmental control parameters and the physiological indicators, which cannot be fitted by the traditional linear model, and there is a lack of adaptive optimization mechanism based on real-time data. Therefore, at present, a more intelligent and efficient pig house environment parameter optimization management technical solution is needed to solve the above problems. SUMMARY

[0004] The purpose of the present application is to provide a pig house environment regulation optimization method and system to solve the problems raised in the background.

[0005] In order to solve the above technical problems, the present application provides a pig house environment regulation optimization method, which comprises:

[0006] S100, collecting pig house environment parameters and pig physiological parameters through a multi-modal sensor.

[0007] The multi-modal sensor includes a thermal imaging camera, an RGB camera, a depth camera and an acoustic acquisition device. The multi-modal sensor is integrated with a track-type inspection robot to collect data column by column. After processing, the pig house environment parameters and the pig physiological parameters are obtained. This ensures that the data covers the entire pig house and avoids the blind area of manual collection.

[0008] In order to process multi-modal data, a special environmental control host is developed based on a high-performance computing platform. A SQL database is established, and the data collected by the sensor and the robot is imported into the database.

[0009] Based on the multi-source sensor integration system of the track-type inspection robot platform, a pollution prevention structure design scheme is constructed to ensure the stability of the data acquisition system. A gradient control scheme is implemented according to the functional zoning of the pig house to eliminate the spatial monitoring blind area. A differentiated data acquisition strategy is implemented for the core functional areas such as the feeding area and the resting area.

[0010] The environmental parameters include the air speed of the fan and the water curtain opening degree of the spraying device; the physiological parameters include the gathering state of the pig group, and the body surface temperature, respiratory frequency and cough frequency of the live pig.

[0011] The gathering state refers to the density of the live pig per unit area. The respiratory frequency refers to the number of breaths per unit time. The cough frequency refers to the number of cough sound events per unit time.

[0012] The automatic and high-precision collection of the pig house environment and live pig physiological data is realized, and the errors caused by human intervention are reduced, providing a comprehensive and real-time data basis for subsequent behavior recognition and environment optimization.

[0013] The monitoring efficiency and reliability are improved, and the data collection process is adapted to the complex environment of the pig house, thereby supporting early risk warning and precise management.

[0014] S200, a behavior recognition model is constructed, and behavior characteristics are analyzed based on physiological parameters and abnormal areas are divided. Specifically, it includes:

[0015] S201, a behavior recognition model is constructed, and the original data collected by the multi-modal sensor is analyzed to identify the gathering state of the pig group, and the body surface temperature, respiratory frequency and cough frequency of the live pig. The construction of the behavior recognition model includes:

[0016] S2011, the original data collected by the multi-modal sensor is preprocessed. It includes:

[0017] An adaptive light compensation algorithm is used to normalize the brightness of the RGB video frame, eliminating the interference of light fluctuation. Ensure stable image quality.

[0018] Based on the live pig skeleton key point detection model, the individual contour is extracted through the pose estimation algorithm, and the contour interpolation reconstruction is performed on the overlapping and occluded area. The occlusion problem when the pig group is crowded is solved, and the individual recognition accuracy is improved.

[0019] The thermal imaging data performs environmental thermal radiation compensation, separates the environmental heat source and the real temperature of the pig body surface through background temperature field modeling. Reduce misjudgment.

[0020] The acoustic signal is subjected to frequency domain noise reduction processing, and the characteristic voiceprint in the 300Hz-3kHz frequency band is extracted. Filter out low-frequency noise and extract characteristic voiceprint for cough event recognition.

[0021] S2012, a behavior characteristic recognition engine is constructed. It includes:

[0022] Gathering state recognition: based on the depth sensor point cloud data, the density value of the live pig per unit area is calculated, and when the density value exceeds the set threshold, the gathering state flag is triggered.

[0023] Body surface temperature inversion: locate the ear root and groin anti-pollution areas of live pigs in thermal imaging data, and output the individual core body surface temperature using a regional temperature weighting algorithm.

[0024] Respiratory frequency detection: capture the temperature fluctuation period of the nostril area of live pigs through thermal imaging sequences, and calculate the respiratory rate per minute by combining chest movement optical flow analysis.

[0025] Cough frequency statistics: build a voiceprint feature matching library, and record an effective cough event when the similarity between the acoustic signal and the cough feature template is greater than the threshold.

[0026] S2013, deploy incremental learning mechanism, including:

[0027] The initial model is trained based on the standard data set of clean pigs; real-time acquisition of dirty environment and group superposition scene data, and generation of enhanced samples through generative adversarial network. Model parameter fine-tuning is performed every 24 hours to update the weight matrix.

[0028] S202, obtain environmental parameters at historical time, analyze the airflow coverage corresponding to different wind speeds, and the spray mapping surface corresponding to different water curtain opening degrees, so as to set the controllable area. Specifically, it includes:

[0029] S2021, obtain environmental parameters at historical time, including the wind speed of the fan at different times, and the water curtain opening degree of the spraying equipment.

[0030] S2022, adopt computational fluid dynamics (CFD) simulation model to build pig house space grid, input fan position and wind speed, generate airflow velocity distribution cloud map corresponding to different wind speeds, and analyze to obtain efficiency area . Specifically, it includes:

[0031] Analyze the wind speed corresponding to each airflow velocity distribution cloud map , set the reference rate and the effective rate , so that they meet . The wind speed is multiplied by and respectively to obtain the reference wind speed and the effective wind speed .

[0032] The continuous area of airflow velocity greater than the reference wind speed in the airflow velocity distribution cloud map is taken as the reference area, and the continuous area of airflow velocity greater than the effective wind speed is taken as the effective area.

[0033] Divide the reference area by the effective area in the airflow velocity distribution cloud map to obtain the efficiency ratio. Take the reference area in the airflow velocity distribution cloud map with the highest efficiency ratio as the efficiency area .

[0034] S2023, establish a spray water droplet motion trajectory model, input water curtain opening degree, simulate water droplet diffusion range, generate spray intensity distribution thermal map corresponding to different water curtain opening degree, analyze to obtain efficiency area . Specifically includes:

[0035] analyze the maximum spray intensity in each spray intensity distribution thermal map , set the reference rate and effective rate , so as to meet . The maximum spray intensity is multiplied by and respectively, to obtain the reference intensity and effective intensity .

[0036] The continuous area of the spray intensity distribution thermal map with spray intensity greater than the reference intensity is regarded as the reference area, and the continuous area of the spray intensity distribution thermal map with spray intensity greater than the effective intensity is regarded as the effective area.

[0037] The area of the reference area in the spray intensity distribution thermal map is divided by the area of the effective area, to obtain the efficiency ratio. The area of the reference area in the spray intensity distribution thermal map with the highest efficiency ratio is regarded as the efficiency area .

[0038] S2024, in the efficiency area and , select the smallest efficiency area as the controllable area.

[0039] S203, analyze the aggregation state of the pig group, and divide the reference area according to the controllable area. Calculate the anomaly index of each reference area respectively, and regard the reference area with anomaly index greater than the threshold value as the abnormal area. Specifically includes:

[0040] S2031, analyze the video image in the original data, connect the two points farthest from each other on each pig to form a line segment, establish a square with the line segment as the diagonal, and select the area of the largest square as the standard area.

[0041] S2032, divide the pig house into grid areas, and the area of each grid is the standard area. Mark the grid where the pig exists, and combine the adjacent marked grids into a reference area, and the area of each reference area is less than or equal to the controllable area.

[0042] S2033, analyze the physiological parameters of each pig in the reference area, and calculate the anomaly index of each reference area by substituting the formula. Regard the reference area with anomaly index greater than the threshold value as the abnormal area. The anomaly index formula is as follows:

[0043] ;

[0044] Where, is a constant greater than 1, and are the average surface temperature and average respiratory rate of pigs in the reference area, and are the average respiratory rate and average body surface temperature of all pigs in the reference area, respectively. The highest coughing frequency among all pigs in the reference area, The maximum coughing frequency allowed under the set health status. is a constant, is the normal pig density per unit area, It is the pig density per unit area in the reference area.

[0045] When the abnormality index exceeds the preset threshold, the reference area is marked as an "abnormal area" and subsequent optimization is triggered. The formula design takes into account both the relative abnormality of the group and the absolute health threshold to avoid misjudgment of a single indicator.

[0046] The algorithm accurately detects abnormal pig behavior and environmental issues, providing early risk warnings and helping optimize resource allocation. This reduces pig health risks, improves the intelligence level of breeding management, and enables targeted intervention by demarcating abnormal areas.

[0047] S300: Establish a physiological environmental control model and set the best optimization plan for each abnormal area. Specifically including:

[0048] S301: Count the number of pigs (s) within the abnormal area (ABN) and create a training set for each pig. Obtain physiological and environmental parameters at the same historical time, classify the pigs according to their type, and place each parameter in the training set for the corresponding pig.

[0049] Each data in the training set is a physiological parameter and environmental parameter at the same time. The physiological parameters refer to the body surface temperature, respiratory rate and coughing frequency of the corresponding pig. The environmental parameters refer to the fan speed acting on the corresponding pig at the same time and the water curtain opening of the spraying equipment.

[0050] S302: Establish a physiological and environmental control model, and use a deep learning algorithm to fit each training set separately to obtain a nonlinear mapping relationship between the physiological parameters and environmental parameters of each pig.

[0051] S303. Set value ranges for each physiological parameter, input these value ranges into the physiological environmental control model, and obtain the value ranges of each environmental parameter for each pig.

[0052] The physiological parameter value interval is set by the staff in advance, referring to the physiological parameter distribution range of the live pigs in a healthy state. The physiological environment control model analyzes the two endpoints of the value interval according to the mapping relationship, and the two results are output as the value interval of the environmental parameter.

[0053] S304, all value intervals are classified according to whether the environmental parameters are the same, and the overlapping interval after the intersection of all value intervals under the same environmental parameter is analyzed as the reference interval of the corresponding environmental parameter.

[0054] S305, n optimization schemes are established for the abnormal area ABN according to all environmental parameters, and the value of each environmental parameter in each optimization scheme is in the corresponding reference interval, and the values of all environmental parameters in different optimization schemes are not completely the same.

[0055] S306, analyze the values of each environmental parameter in the optimization scheme, calculate the comfort index of each optimization scheme respectively, and select the optimization scheme with the largest comfort index as the best optimization scheme of the abnormal area ABN.

[0056] The values of each environmental parameter in the optimization scheme are input into the physiological environment control model to obtain the predicted values of each physiological parameter of each live pig. The comfort index is calculated by substituting the formula

[0057]

[0058] In the formula, is a constant greater than 1, is the maximum allowed cough frequency in the set healthy state, is the maximum predicted value of the cough frequency of all live pigs, is a constant, and are the respiratory rate and body surface temperature in the set healthy state respectively. and are the predicted values of the body surface temperature and respiratory rate of the first live pig respectively.

[0059] For each abnormal area, the comfort index of the multiple optimization schemes generated is calculated. The scheme corresponding to the maximum value is selected to ensure that the cough is suppressed first and the respiratory rate and body temperature tend to be healthy.

[0060] The abnormal index realizes early risk positioning and improves monitoring efficiency. The comfort index quantifies the prediction effect, drives the adaptive adjustment of the environmental control parameters, and improves the breeding benefit.

[0061] Similarly, the best optimization scheme is planned for each abnormal area.

[0062] Both the abnormal index and the comfort index are dimensionless quantities.​​​

[0063] Through model-driven optimization, the environmental parameters are accurately adapted to the needs of the abnormal area, maximizing the comfort of pigs and reducing the incidence of disease. This improves the adaptive management efficiency of the pig house environment, optimizes resource utilization, and improves overall breeding benefits.

[0064] S400, generate environmental control instructions according to the optimal optimization scheme and control the equipment to execute.

[0065] The optimal optimization scheme for each abnormal area is obtained, and the environmental control instructions are generated according to the values of the environmental parameters in the optimal optimization scheme, and the fan speed and water curtain opening of the spraying equipment are controlled according to the environmental control instructions.

[0066] A fast and accurate response mechanism is achieved, ensuring that the environmental optimization scheme takes effect immediately and reducing the delay of manual operation. The pig house microclimate is optimized, resource utilization efficiency is improved, and the stability and sustainability of breeding production are enhanced.

[0067] The present application also provides a pig house environment regulation and optimization system, which comprises an intelligent sensing module, an environmental analysis module, a parameter optimization module and a control management module.

[0068] The intelligent sensing module is used to collect pig house environmental parameters and pig physiological parameters through multi-modal sensors.

[0069] The multi-modal sensor is integrated with the track inspection robot to collect data by row, and the original data is processed to obtain pig house environmental parameters and pig physiological parameters. The specific collection process involves adaptive light compensation, skeleton key point detection, environmental thermal radiation compensation and acoustic frequency domain noise reduction technology to ensure data quality.

[0070] The implementation of automatic and high-precision multi-modal data collection reduces human intervention errors and provides a comprehensive and real-time data basis for subsequent behavior recognition and environmental optimization, improving the efficiency and reliability of pig house monitoring.

[0071] The environmental analysis module is used to build a behavior recognition model, analyze behavior characteristics based on physiological parameters and divide abnormal areas.

[0072] The behavior recognition model is built, including data preprocessing and behavior feature recognition, and an incremental learning mechanism is deployed. Then, combined with historical environmental parameter analysis, the controllable area is calculated, the reference area is divided and the abnormal index is calculated to identify abnormal areas.

[0073] Through advanced algorithms, abnormal behaviors of pigs and environmental problems are accurately detected, early risk warning is realized, resource allocation is optimized, pig health risks are reduced, and the intelligent level of breeding management is improved.

[0074] The parameter optimization module is used to establish a physiological environmental control model to set the optimal optimization scheme for each abnormal area.

[0075] A physiological control model is established, a deep learning algorithm is used to fit the training set of each pig, and a nonlinear mapping is derived; a health value interval is set for the physiological parameters, a reference interval of the environmental parameters is obtained by inputting the model, a plurality of optimization schemes are generated, and the best optimization scheme is selected through the comfort index formula.

[0076] Through model-driven optimization, the environmental parameters are accurately adapted to the needs of the abnormal area, the pig comfort is maximized, the disease incidence is reduced, and the adaptive management efficiency and overall breeding benefit of the pig house environment are improved.

[0077] The control management module is used to generate the environmental control instructions according to the best optimization scheme and control the equipment to execute.

[0078] Based on the best optimization scheme output by the parameter optimization module, specific environmental control instructions are generated, and these instructions are executed through the control system to realize the automatic adjustment of the equipment.

[0079] A fast and accurate response mechanism is realized to ensure that the environmental optimization scheme takes effect immediately, reduce the delay of manual operation, optimize the microclimate of the pig house, improve the resource utilization efficiency, and ultimately enhance the stability and sustainability of the breeding production.

[0080] Compared with the prior art, the beneficial effects achieved by the present application are:

[0081] Multi-modal data acquisition advantage: By integrating thermal imaging cameras, RGB cameras, depth cameras and acoustic acquisition devices into the track-type inspection robot, the pig house environment and the physiological parameters of pigs are comprehensively and synchronously collected. This method eliminates the interference of light fluctuations, environmental thermal radiation and acoustic noise, provides a high-precision data basis, and avoids the coverage deficiency and error accumulation problems caused by single sensors in the prior art.

[0082] Intelligent behavior recognition and anomaly detection advantage: A behavior recognition model is constructed, advanced algorithms such as adaptive light compensation, skeleton key point detection, thermal radiation compensation and frequency domain noise reduction are used to accurately identify the pig group aggregation state, body surface temperature, respiratory rate and cough frequency. Through the incremental learning mechanism, the model is continuously optimized to adapt to the dirty environment and group superposition scene, improve the accuracy and robustness of abnormal area division, and overcome the failure risk of existing methods in dynamic changing scenes.

[0083] Personalized environmental optimization advantage: A physiological control model is established, and a deep learning algorithm is used to fit the nonlinear mapping relationship between physiological parameters and environmental parameters for each pig to generate customized optimization schemes. Based on the comfort index, the best scheme is dynamically selected to implement precise regulation for the abnormal area, such as fan speed and water curtain opening degree adjustment, to ensure pig comfort, and solve the resource waste and poor effect problems caused by the one-size-fits-all environmental control in the prior art.

[0084] System integration and automation advantages: The whole scheme adopts modular design, including intelligent sensing, environment analysis, parameter optimization and control management module, realizes the whole process automation from data acquisition to instruction execution. Through the track inspection robot and real-time environmental control instruction, reduce manual intervention, improve management efficiency and response speed, better than the existing segmented system in the deficiency of cooperation and real-time. BRIEF DESCRIPTION OF DRAWINGS

[0085] The accompanying drawings are included to provide a further understanding of the application, and constitute a part of the specification, illustrate the application, and are used to explain the application together with the embodiments of the application, and do not constitute a limitation on the application. In the drawings:

[0086] Figure 1 is a flowchart of the pig house environment regulation optimization method of the application;

[0087] Figure 2 is a structural schematic diagram of the pig house environment regulation optimization system of the application. DETAILED DESCRIPTION

[0088] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0089] Please refer to Figure 1 , the application provides a pig house environment regulation optimization method, comprising:

[0090] S100, collecting pig house environment parameters and pig physiological parameters through multi-modal sensors.

[0091] The multi-modal sensor includes a thermal imaging camera, an RGB camera, a depth camera and an acoustic acquisition device. The track inspection robot integrates the multi-modal sensor to collect data column by column, and obtains the pig house environment parameters and the pig physiological parameters after processing. It ensures that the data covers the whole pig house and avoids the blind area of manual collection.

[0092] In order to process multi-modal data, a special environmental control host is developed based on a high-performance computing platform. A SQL database is established, and the data collected by the sensor and the robot is imported into the database.

[0093] A multi-source sensor integration system based on a trackable inspection robot platform is constructed, a pollution prevention structure design scheme is constructed to ensure the stability of the data acquisition system, a gradient control scheme is implemented according to the functional partitioning of the pig house to eliminate the spatial monitoring blind area, and a differentiated data acquisition strategy is implemented for core functional areas such as the feeding area and the resting area.

[0094] The environmental parameters include the wind speed of the fan and the water curtain opening of the spraying equipment; the physiological parameters include the gathering state of the pig group, and the body surface temperature, respiratory frequency and cough frequency of the pigs.

[0095] The gathering state refers to the pig density per unit area. The respiratory frequency refers to the number of breaths per unit time. The cough frequency refers to the number of cough sound events per unit time.

[0096] The automatic and high-precision collection of pig house environment and pig physiological data is realized, and the errors caused by human intervention are reduced, providing a comprehensive and real-time data basis for subsequent behavior recognition and environment optimization.

[0097] The monitoring efficiency and reliability are improved, and the data acquisition process is adapted to the complex environment of the pig house (such as changes in light and noise interference), thereby supporting early risk warning and precise management.

[0098] S200, a behavior recognition model is constructed, and behavior characteristics are analyzed based on physiological parameters to divide abnormal areas. Specifically, it includes:

[0099] S201, a behavior recognition model is constructed, and the original data collected by the multi-modal sensor is analyzed to identify the gathering state of the pig group, and the body surface temperature, respiratory frequency and cough frequency of the pigs. The construction of the behavior recognition model includes:

[0100] S2011, the original data collected by the multi-modal sensor is preprocessed. It includes:

[0101] An adaptive light compensation algorithm is used to normalize the brightness of the RGB video frame, eliminating the interference of light fluctuation. Ensure stable image quality.

[0102] Based on the pig skeleton key point detection model, the individual contour is extracted through the pose estimation algorithm, and the contour interpolation reconstruction is implemented for the overlapping and occluded area. The occlusion problem when the pig group is crowded is solved, and the individual recognition accuracy is improved.

[0103] The thermal imaging data performs environmental thermal radiation compensation, separates the environmental heat source from the real temperature of the pig body surface through background temperature field modeling. Reduce misjudgment.

[0104] The acoustic signal is subjected to frequency domain noise reduction processing, and the characteristic voiceprint in the 300Hz-3kHz frequency band is extracted. Low-frequency noise (such as fan sound) is filtered out, and the characteristic voiceprint is extracted for cough event recognition.

[0105] S2012, build behavior feature recognition engine. Including:

[0106] Aggregation state recognition: based on depth sensor point cloud data, calculate the density value of pigs per unit area, when the density value exceeds the set threshold, trigger the aggregation state flag.

[0107] Body surface temperature inversion: locate the ear root and groin anti-pollution area of pigs in thermal imaging data, and output the individual core body surface temperature using regional temperature weighting algorithm.

[0108] Respiratory rate detection: capture the temperature fluctuation period of the nostril area of pigs through thermal imaging sequence, combine with chest movement optical flow analysis, calculate the respiratory rate per minute.

[0109] Cough frequency statistics: build a voiceprint feature matching library, when the similarity of acoustic signal and cough feature template is greater than the threshold, record it as an effective cough event.

[0110] S2013, deploy incremental learning mechanism, including:

[0111] The initial model is trained based on the standard data set of clean pig groups; real-time acquisition of dirty environment and group superposition scene data, generate enhanced samples through generative adversarial network. Model parameter fine-tuning is performed every 24 hours, and the weight matrix is updated.

[0112] S202, obtain environmental parameters at historical time, analyze the airflow coverage corresponding to different wind speeds, and the spray mapping surface corresponding to different water curtain opening degrees, so as to set the controllable area. Specifically including:

[0113] S2021, obtain environmental parameters at historical time, including the wind speed of the fan at different times, and the water curtain opening degree of the spraying equipment.

[0114] S2022, adopt computational fluid dynamics (CFD) simulation model to build pig house space grid, input fan position and wind speed, generate airflow velocity distribution cloud map corresponding to different wind speeds, and analyze to obtain effective area . Specifically including:

[0115] Analyze the wind speed corresponding to each airflow velocity distribution cloud map , set the reference rate and the effective rate , so that they meet . The wind speed is multiplied by and respectively to obtain the reference wind speed and the effective wind speed .

[0116] The continuous area of the airflow velocity distribution cloud map greater than the reference wind speed is taken as the reference area, and the continuous area greater than the effective wind speed is taken as the effective area.

[0117] The reference area in the airflow velocity distribution cloud map is divided by the effective area to obtain the efficiency ratio. The reference area in the airflow velocity distribution cloud map with the highest efficiency ratio is taken as the efficiency area .

[0118] S2023, a spray water droplet motion trajectory model is established, the water curtain opening degree is input, the water droplet diffusion range is simulated, the spray intensity distribution thermal map corresponding to different water curtain opening degrees is generated, and the efficiency area is obtained by analysis . Specifically, it includes:

[0119] The maximum spray intensity in each spray intensity distribution thermal map is analyzed , the reference rate and the effective rate are set to meet . The maximum spray intensity is multiplied by and respectively to obtain the reference intensity and the effective intensity .

[0120] The continuous area of the spray intensity distribution thermal map greater than the reference intensity is taken as the reference area, and the continuous area greater than the effective intensity is taken as the effective area.

[0121] The reference area in the spray intensity distribution thermal map is divided by the effective area to obtain the efficiency ratio. The reference area in the spray intensity distribution thermal map with the highest efficiency ratio is taken as the efficiency area .

[0122] S2024, in the efficiency area and , the smallest efficiency area is selected as the controllable area.

[0123] S203, analyze the aggregation state of the pig group, and divide the reference area according to the controllable area. Calculate the anomaly index of each reference area, and take the reference area with an anomaly index greater than the threshold value as the abnormal area. Specifically, it includes:

[0124] S2031, analyze the video images in the original data, connect the two points farthest from each other on each pig to form a line segment, and take the line segment as the diagonal to establish a square. The area of the largest square is taken as the standard area.

[0125] S2032, divide the pig house into grids, and the area of each grid is a standard area. Mark the grid where the live pigs exist, and combine adjacent marked grids into a reference area, and the area of each reference area is less than or equal to a controllable area.

[0126] S2033, analyze the physiological parameters of each live pig in the reference area, and calculate the abnormal index of each reference area by substituting the formula. The reference area with an abnormal index greater than a threshold value is regarded as an abnormal area. The abnormal index formula As follows:

[0127] ;

[0128] In the formula, is a constant greater than 1, and are the average body surface temperature and the average respiratory rate of the live pigs in the reference area, and are the average respiratory rate and the average body surface temperature of the live pigs in all reference areas. is the highest cough frequency among all live pigs in the reference area, is the maximum cough frequency allowed under the set healthy state. is a constant, is the set normal live pig density per unit area, is the live pig density per unit area in the reference area.

[0129] When the abnormal index exceeds the preset threshold value, the reference area is marked as an "abnormal area" and triggers subsequent optimization. The formula takes into account both the relative abnormality of the group and the absolute health threshold, avoiding misjudgment by a single indicator.

[0130] By accurately detecting abnormal behaviors of the pig group and environmental problems through algorithms, early risk warnings (such as diseases or stress) are achieved, helping to optimize resource allocation (such as directional adjustment of fan and spraying equipment). The health risks of live pigs are reduced, and the intelligent level of breeding management is improved. Through the division of abnormal areas, targeted intervention is realized.

[0131] S300, establish a physiological environment control model, and set the best optimization scheme for each abnormal area. Specifically including:

[0132] S301, count the number of live pigs in the abnormal area ABN, and establish a training set for each live pig. Obtain the physiological parameters and environmental parameters at the same time in history, and classify them according to the corresponding live pigs. Each type of parameter is placed in the training set of the corresponding live pig.

[0133] Each data in the training set is the physiological parameter and environmental parameter at the same time. The physiological parameter refers to the body surface temperature, respiratory rate, and cough frequency of the corresponding live pig, and the environmental parameter refers to the fan speed and water curtain opening degree of the spraying equipment acting on the corresponding live pig at the same time.

[0134] S302, a physiological control model is established, and a deep learning algorithm is used to fit each training set to obtain a nonlinear mapping relationship between the physiological parameters and the environmental parameters of each pig.

[0135] S303, a value interval is set for each physiological parameter, and the value intervals are input into the physiological control model to obtain the value interval of each environmental parameter for each pig.

[0136] The value interval of the physiological parameter is set by the staff in advance, referring to the physiological parameter distribution range of the pig in a healthy state. The physiological control model analyzes the two endpoints of the value interval according to the mapping relationship, and the two results are output as the value interval of the environmental parameter.

[0137] S304, all value intervals are classified according to whether the environmental parameters are the same, and the overlapping interval after the intersection of all value intervals under the same environmental parameter is analyzed as the reference interval of the corresponding environmental parameter.

[0138] S305, n optimization schemes are established for the abnormal zone ABN according to all environmental parameters, and the values of the environmental parameters in each optimization scheme are in the corresponding reference interval, and the values of all environmental parameters in different optimization schemes are not completely the same.

[0139] S306, the values of the environmental parameters in the optimization scheme are analyzed, the comfort index of each optimization scheme is calculated, and the optimization scheme with the largest comfort index is selected as the best optimization scheme of the abnormal zone ABN.

[0140] The values of the environmental parameters in the optimization scheme are input into the physiological control model to obtain the predicted values of each physiological parameter for each pig. The comfort index is calculated by substituting the formula

[0141]

[0142] In the formula, is a constant greater than 1, is the maximum cough frequency allowed in the set healthy state, is the maximum predicted value of the cough frequency of all pigs, is a constant, and are the respiratory rate and body temperature in the set healthy state, respectively. and are the predicted values of the body temperature and respiratory rate of the first pig, respectively.

[0143] ​​​A plurality of optimization schemes (such as different wind speed and water curtain opening degree combinations) are generated for each abnormal area, and the comfort index is calculated. The scheme corresponding to the maximum value is selected to ensure that cough is suppressed as a priority (for infectious disease prevention and control) and that respiration and body temperature approach healthy levels (for thermal comfort).

[0144] The abnormal index enables early risk positioning (such as a sudden increase in cough frequency) and improves monitoring efficiency. The comfort index quantifies the prediction effect and drives adaptive adjustment of environmental control parameters (such as increasing wind speed in high temperatures), thereby improving breeding efficiency.

[0145] Similarly, the best optimization scheme is planned for each abnormal area.

[0146] Both the abnormal index and the comfort index are dimensionless quantities.

[0147] Through model-driven optimization, environmental parameters (such as wind speed and water curtain opening degree) are precisely adapted to the needs of abnormal areas, maximizing pig comfort and reducing disease incidence. This improves the efficiency of adaptive management of pig house environments, optimizes resource utilization, and improves overall breeding efficiency.

[0148] S400, generate environmental control instructions based on the best optimization scheme and control the equipment to execute.

[0149] The best optimization scheme for each abnormal area is obtained, and environmental control instructions are generated based on the values of the environmental parameters in the best optimization scheme. The wind speed of the fan and the water curtain opening degree of the spraying equipment are controlled to execute according to the environmental control instructions.

[0150] A fast and accurate response mechanism is achieved, ensuring that environmental optimization schemes take effect immediately and reducing the delay of manual operations. The pig house microclimate (such as temperature and humidity) is optimized, improving resource utilization efficiency and enhancing the stability and sustainability of breeding production.

[0151] See Figure 2 The present application also provides a pig house environment regulation and optimization system, which includes an intelligent sensing module, an environmental analysis module, a parameter optimization module, and a control management module.

[0152] The intelligent sensing module is used to collect pig house environmental parameters and pig physiological parameters through multi-modal sensors.

[0153] Through the integration of multi-modal sensors (including thermal imaging cameras, RGB cameras, depth cameras, and acoustic acquisition devices) on a track-type inspection robot, data is collected row by row, and raw data is processed to obtain pig house environmental parameters (such as fan wind speed and spraying equipment water curtain opening degree) and pig physiological parameters (such as pig group aggregation state, pig body surface temperature, respiratory rate, and cough frequency). The specific collection process involves adaptive illumination compensation, skeletal key point detection, environmental thermal radiation compensation, and acoustic frequency domain noise reduction techniques to ensure data quality.

[0154] The automatic and high-precision multi-modal data acquisition is realized, the human intervention error is reduced, a comprehensive and real-time data basis is provided for subsequent behavior recognition and environment optimization, and the efficiency and reliability of the pig house monitoring are improved.

[0155] The environment analysis module is used to build a behavior recognition model, analyze behavior characteristics based on physiological parameters, and divide abnormal areas.

[0156] The behavior recognition model is built, including data preprocessing (such as brightness normalization of RGB video frames, environmental thermal radiation compensation of thermal imaging, and frequency domain noise reduction of acoustic signals) and behavior feature recognition (such as calculating the aggregation state based on point cloud data, analyzing the respiratory rate of thermal imaging sequence, and matching the acoustic frequency to calculate the cough frequency), and deploying an incremental learning mechanism (fine-tuning the model every 24 hours). Subsequently, combined with the historical environmental parameter analysis, the controllable area (the effective area is calculated by CFD simulation and spraying model) is divided into reference areas and the abnormal index (based on the physiological parameter formula) is calculated to identify abnormal areas.

[0157] By using advanced algorithms to accurately detect abnormal behaviors of pig groups and environmental problems, early risk warning is realized, which helps to optimize resource allocation, reduce the health risks of live pigs, and improve the intelligent level of breeding management.

[0158] The parameter optimization module is used to establish a physiological environmental control model to set the best optimization scheme for each abnormal area.

[0159] The physiological environmental control model is established, the training set of each live pig (including the mapping relationship of historical physiological parameters and environmental parameters) is fitted using deep learning algorithm, and the nonlinear mapping is derived; the health value interval of physiological parameters is set, the reference interval of environmental parameters is input into the model, multiple optimization schemes (each environmental parameter value is in the reference interval) are generated, and the best optimization scheme is selected through the comfort index formula (based on the predicted physiological parameters).

[0160] Through model-driven optimization, the environmental parameters (such as wind speed and water curtain opening) are accurately adapted to the needs of abnormal areas, the comfort of live pigs is maximized, the disease incidence is reduced, and the adaptive management efficiency of pig house environment and the overall breeding benefit are improved.

[0161] The control management module is used to generate environmental control instructions based on the best optimization scheme and control the equipment to execute.

[0162] Based on the best optimization scheme output by the parameter optimization module, specific environmental control instructions (such as adjusting the wind speed of the fan and the water curtain opening of the spraying equipment) are generated, and these instructions are executed through the control system to realize the automatic adjustment of the equipment.

[0163] A fast and accurate response mechanism is realized to ensure that the environmental optimization scheme takes effect immediately, reduce the delay of manual operation, optimize the pig house microclimate, improve the resource utilization efficiency, and ultimately enhance the stability and sustainability of breeding production.

[0164] In embodiment 1, it is assumed that there are several pigs in the reference area A1, the average body surface temperature of the pigs is 39℃, the average respiratory rate is 18 times per minute, the maximum cough frequency is 3 times per hour, and the pig density per unit area is 0.6 pigs / m².

[0165] When the constant is 2, the average respiratory rate of the pigs in all reference areas is 15 times per minute, the average body surface temperature is 38℃, the maximum cough frequency allowed under the set health state is 5 times per hour, the constant is 1, and the normal pig density per unit area is set to 1 pig / m², the abnormal index of the reference area A1 is calculated by substituting the formula:

[0166] Reference area A1: ;

[0167] The abnormal index of the reference area A1 is 1.49.

[0168] It should be noted that the relational terms such as first and second are used only to differentiate one entity or action from another, and do not necessarily require or imply that these entities or actions exist in any actual relationship or order. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0169] Finally, it should be noted that the above description is only for the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent replacements to some technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for optimizing piggery environment control, characterized by: The method includes: S100, collecting pig house environmental parameters and pig physiological parameters through multimodal sensors; S200, constructing a behavior recognition model, analyzing behavior characteristics based on physiological parameters and dividing abnormal areas; S300, establish a physiological environment control model and set the best optimization plan for each abnormal area; S400: Generate environmental control instructions according to the best optimization solution and control the equipment to execute them.

2. The method for optimizing the piggery environment control according to claim 1, characterized in that: In the S100, multimodal sensors include thermal imaging cameras, RGB cameras, depth cameras, and acoustic acquisition devices. The track-mounted inspection robot integrates multimodal sensors to collect data pen by pen, and after processing, obtains pig house environmental parameters and pig physiological parameters; Environmental parameters include the wind speed of the fan and the opening of the water curtain of the spraying equipment; physiological parameters include the aggregation state of the pig group, as well as the pig's body temperature, respiratory rate and coughing frequency; Aggregation status refers to the density of pigs per unit area; Respiratory frequency refers to the number of breaths per unit time; cough frequency refers to the number of cough events per unit time.

3. The method for optimizing the piggery environment control according to claim 2, characterized in that: S200 includes: S201. Build a behavior recognition model to identify the aggregation state of the pig herd, as well as the pigs' body surface temperature, respiratory rate, and coughing frequency by analyzing the raw data collected by the multimodal sensor; S202: Acquire historical environmental parameters, analyze airflow coverage corresponding to different wind speeds, and spray mapping surfaces corresponding to different water curtain openings, thereby setting a controllable area; S203, analyzing the aggregation state of the pig herd, dividing the reference area according to the controllable area; calculating the abnormal index of each reference area respectively, and taking the reference area with an abnormal index greater than a threshold as an abnormal area.

4. The method for optimizing the piggery environment control according to claim 3, characterized in that: In S201, the construction of the behavior recognition model includes: S2011. Preprocessing the raw data collected by the multimodal sensor; including: Adopting adaptive illumination compensation algorithm to normalize the brightness of RGB video frames and eliminate the interference of illumination fluctuations; Based on the pig skeleton key point detection model, the pose estimation algorithm is used to extract the individual body contours, and the body contours of the overlapping occluded areas are interpolated and reconstructed. Thermal imaging data performs environmental thermal radiation compensation, and separates the environmental heat source from the actual pig surface temperature through background temperature field modeling; The acoustic signal is processed in the frequency domain to reduce noise and extract characteristic voiceprints in the 300Hz-3kHz frequency band; S2012. Build a behavioral feature recognition engine; including: Aggregation status recognition: Based on the depth sensor point cloud data, the pig density value per unit area is calculated, and the aggregation status flag is triggered when the density value exceeds the set threshold; Surface temperature inversion: Locate the pig's ear and groin anti-contamination areas in thermal imaging data and use a regional temperature weighted algorithm to output the individual core surface temperature; Respiratory rate detection: The temperature fluctuation cycle of the pig's nostril area is captured through thermal imaging sequences, and the number of respirations per minute is calculated by combining chest cavity motion optical flow analysis; Cough frequency statistics: Build a voiceprint feature matching library and record a valid cough event when the similarity between the acoustic signal and the cough feature template is greater than a threshold; S2013. Deploy incremental learning mechanisms, including: The initial model is trained based on a standard dataset of clean pig herds; real-time data on dirty environments and overlapping herds is collected, and enhanced samples are generated through a generative adversarial network; model parameter fine-tuning is performed every 24 hours to update the weight matrix.

5. The method for optimizing the piggery environment control according to claim 3, characterized in that: S202 includes: S2021. Obtain historical environmental parameters, including wind speed of fans at different times and water curtain opening of spray equipment; S2022. Use the computational fluid dynamics (CFD) simulation model to construct the pig house space grid, input the fan position and wind speed, generate the air flow velocity distribution cloud map corresponding to different wind speeds, and analyze and obtain the effective area. ; Specifically include: Analyze the wind speed corresponding to each airflow velocity distribution cloud map , set the base rate and efficiency , so that it satisfies Wind speed Respectively and Multiply them to get the reference wind speed and effective wind speed ; The continuous area where the airflow velocity is greater than the reference wind speed in the airflow velocity distribution cloud map is regarded as the reference area, and the continuous area where the airflow velocity is greater than the effective wind speed is regarded as the effective area; The efficiency ratio is obtained by dividing the area of ​​the reference area in the air velocity distribution cloud map by the area of ​​the effective area; the area of ​​the reference area in the air velocity distribution cloud map with the highest efficiency ratio is taken as the effective area. ; S2023. Establish a spray droplet motion trajectory model, simulate the water droplet diffusion range after inputting the water curtain opening, generate a spray intensity distribution heat map corresponding to different water curtain openings, and analyze and obtain the effective area ; Specifically include: Analyze the maximum spray intensity in each spray intensity distribution heat map , set the base rate and efficiency , so that it satisfies ; Maximum spray intensity Respectively and Multiply them to get the baseline intensity and effective strength ; The continuous area where the spray intensity is greater than the reference intensity in the spray intensity distribution thermodynamic map is taken as the reference area, and the continuous area where the spray intensity is greater than the effective intensity is taken as the effective area; The efficiency ratio is obtained by dividing the area of ​​the reference area in the spray intensity distribution thermodynamic map by the area of ​​the effective area. The area of ​​the reference area in the spray intensity distribution thermodynamic map with the highest efficiency ratio is taken as the effective area. ; S2024, in the effective area and In the equation, the smallest effective area is selected as the controllable area.

6. The method for optimizing the piggery environment control according to claim 3, characterized in that: S203 includes: S2031. Analyze the video images in the original data, connect the two farthest points on each pig as a line segment, use the line segment as a diagonal to create a square, and select the area of ​​the largest square as the standard area; S2032. Divide the piggery into grid areas, with the area of ​​each grid being the standard area; mark grids where pigs are present, and combine adjacent marked grids into reference areas, with the area of ​​each reference area being less than or equal to the controllable area; S2033. Analyze the physiological parameters of each pig in the reference area and substitute them into the formula to calculate the abnormal index of each reference area; the reference area with an abnormal index greater than the threshold is regarded as an abnormal area; the abnormal index formula is as follows: ； Where, is a constant greater than 1, and are the average surface temperature and average respiratory rate of pigs in the reference area, and are the average respiratory rate and average body surface temperature of all pigs in the reference area; The highest coughing frequency among all pigs in the reference area, The maximum coughing frequency allowed under the set health status; is a constant, is the normal pig density per unit area, It is the pig density per unit area in the reference area.

7. The method for optimizing the piggery environment control according to claim 3, characterized in that: S300 includes: S301. Count the number of pigs s in the abnormal area ABN and create a training set for each pig. Obtain physiological parameters and environmental parameters at the same time in history, classify the pigs according to their category, and put each type of parameter into the training set of the corresponding pig. S302: Establish a physiological and environmental control model, and use a deep learning algorithm to fit each training set to obtain a nonlinear mapping relationship between the physiological parameters and environmental parameters of each pig; S303, setting value intervals for each physiological parameter, inputting these value intervals into the physiological environmental control model, and obtaining the value intervals of each environmental parameter for each pig; S304: All value intervals are classified according to whether the environmental parameters are the same, and the overlapping intervals after the intersection of all value intervals under the same environmental parameters are analyzed as the reference interval of the corresponding environmental parameters; S305. Establish n optimization schemes for the abnormal area ABN based on all environmental parameters. The values ​​of the environmental parameters in each optimization scheme are within the corresponding reference range. The values ​​of all environmental parameters in different optimization schemes are not exactly the same. S306, analyzing the values ​​of various environmental parameters in the optimization scheme, calculating the comfort index of each optimization scheme respectively, and selecting the optimization scheme with the largest comfort index as the best optimization scheme for the abnormal area ABN; And so on, the best optimization plan is planned for each abnormal area.

8. The method for optimizing the piggery environment control according to claim 7, characterized in that: In S306, the values ​​of each environmental parameter in the optimization scheme are input into the physiological environmental control model to obtain the predicted values ​​of each physiological parameter for each pig; Substitute the formula to calculate the comfort index : ； Where, is a constant greater than 1, The maximum coughing frequency allowed under the set health status, is the maximum predicted value of cough frequency among all pigs, is a constant, and They are respectively the respiratory rate and body surface temperature under the set health status; and Respectively Predicted values ​​of skin temperature and respiratory rate for first-born pigs.

9. The method for optimizing the piggery environment control according to claim 7, characterized in that: In S400, the best optimization scheme for each abnormal area is obtained respectively, and an environmental control instruction is generated according to the values ​​of various environmental parameters in the best optimization scheme, and the wind speed of the fan and the water curtain opening of the spraying equipment are controlled according to the environmental control instruction.

10. Piggery environment control and optimization system, characterized by: The system includes intelligent perception module, environmental analysis module, parameter optimization module and control management module; The intelligent sensing module is used to collect pig house environmental parameters and pig physiological parameters through multimodal sensors; The environmental analysis module is used to build a behavior recognition model, analyze behavioral characteristics based on physiological parameters, and divide abnormal areas; The parameter optimization module is used to establish a physiological environmental control model and set the best optimization plan for each abnormal area; The control management module is used to generate environmental control instructions according to the best optimization plan and control the execution of equipment.

Images