Data monitoring and processing method and system based on Internet of Things livestock breeding early warning platform

Abstract

The invention relates to the technical field of data monitoring, in particular to a data monitoring processing method and system based on an Internet of Things livestock breeding early warning platform. The method comprises the following steps: collecting dust vertical distribution gradient data in real time through a multi-level PM2.5 sensor array, and dynamically cleaning PM2.5 sensors; uploading the dust vertical distribution gradient data to a preset livestock breeding early warning platform; collecting a first target livestock group density distribution diagram and a first livestock group individual displacement vector sequence in real time; constructing a target livestock group kinematic model based on the first target livestock group density distribution map and the first livestock group individual displacement vector sequence; and performing biological disturbance compensation on the dust vertical distribution gradient data according to the target animal group kinematics model to generate first dust vertical distribution gradient data. The problems of baseline drift and compensation lag in existing dust monitoring are solved, and the monitoring precision of the livestock breeding early warning platform is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data monitoring, and particularly to a data monitoring and processing method and system based on an Internet of Things (IoT) livestock breeding warning platform. Background Art

[0002] In the precise environmental control scenario of a closed Hu sheep breeding house, the application of IoT technology provides powerful data support and intelligent management means for the livestock breeding warning platform. By deploying a large number of sensor nodes inside the breeding house, including PM2.5 sensors for monitoring air quality, infrared array sensors for monitoring herd behavior, etc., these sensor nodes can collect various data in the breeding environment in real time and transmit the data to the central control system through wireless communication technology, forming a complete IoT monitoring network. The livestock breeding warning platform, as the core of data processing and analysis, uses these real-time data for environmental monitoring, data analysis, and warning decision-making, providing precise environmental control suggestions for breeding management personnel.

[0003] However, the existing livestock breeding warning platform exposes significant adaptability defects in the dynamic compensation mechanism of dust monitoring data. For example, when the automatic feeding system feeds concentrated feed during the periods of 06:00 - 06:15 and 18:00 - 18:15 every day, corn-soybean meal mixed feed dust with a particle size less than 50 μm forms a concentration gradient distribution in the vertical space. At this time, hydrophobic organic particles accumulate on the surface of the photoelectric detection unit of the PM2.5 sensor at a rate of 3 - 5 μg / cm 2 per minute, causing the baseline value of the sensor to drift upward at a rate of 8 - 12% per hour. The existing compensation algorithms only adopt a zero-point calibration strategy with a fixed period (usually set to 24 hours), failing to recognize the strong spatio-temporal correlation between the feeding operation and the baseline drift, resulting in a compensation lag effect that persists in the critical monitoring window period of 35 - 40 minutes after feeding. This lag seriously affects the real-time monitoring and precise warning capabilities of the livestock breeding warning platform for dust concentration changes, and further affects the precise control of the breeding environment. Summary of the Invention

[0004] Based on this, it is necessary for the present invention to provide a data monitoring and processing method and system based on an IoT livestock breeding warning platform to solve at least one of the above technical problems.

[0005] To achieve the above object, a data monitoring and processing method based on an IoT livestock breeding warning platform includes the following steps:

[0006] Step S1: Real-time collect dust vertical distribution gradient data through a multi-level PM2.5 sensor array, and dynamically clean the PM2.5 sensor; upload the dust vertical distribution gradient data to a preset livestock breeding warning platform;

[0007] Step S2: Collect the first target herd density distribution map and the first herd individual displacement vector sequence in real time; construct a target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence;

[0008] Step S3: Perform bioturbation compensation on the dust vertical distribution gradient data according to the target herd kinematic model to generate the first dust vertical distribution gradient data;

[0009] Step S4: Detect the end signal of the feeding event in real time. When the end signal of the feeding event is detected, perform zonal calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

[0010] Through dynamically cleaning the PM2.5 sensor and collecting the dust vertical distribution gradient data in real time, the present invention effectively solves the problem of baseline drift caused by the accumulation of hydrophobic organic particles on the sensor in the prior art, avoids the compensation lag effect caused by the fixed-period zero calibration strategy, and ensures the accuracy and real-time nature of the dust monitoring data. At the same time, by collecting the herd density distribution and the individual displacement vector sequence to construct a kinematic model and compensating the dust data accordingly, it can accurately eliminate the influence of herd behavior dynamics interference on the dust monitoring data, solves the problem that the traditional compensation model does not consider dynamic air disturbance and mechanical vibration noise, and improves the reliability of the monitoring data. By detecting the end signal of the feeding event in real time and performing zonal calibration, the accuracy of the dust data collected by the livestock breeding warning platform is further optimized.

[0011] Preferably, the sampling frequency adjustment rule of each PM2.5 sensor in the multi-level PM2.5 sensor array in step S1 is specifically:

[0012] Detect the feeding event signal in real time;

[0013] Based on a preset breeding event trigger signal, determine whether the feeding event signal is a feeding period trigger signal;

[0014] If the feeding event signal is not a feeding period trigger signal, identify the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the standard mode, maintain the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the high-frequency monitoring mode, switch the current sampling frequency mode to the standard mode;

[0015] If the feeding event signal is a feeding period trigger signal, switch the current sampling frequency mode of the PM2.5 sensor to the high-frequency monitoring mode.

[0016] By dynamically adjusting the sampling frequency of the multi-level PM2.5 sensor array, the present invention realizes the refined management of dust monitoring in the livestock breeding environment. During the feeding period, the sensor switches to the high-frequency monitoring mode, which can more sensitively capture the rapid changes in dust concentration, ensuring high-precision data acquisition when the dust concentration gradient distribution changes drastically. During the non-feeding period, the sensor maintains the standard mode, which not only saves energy consumption but also avoids unnecessary data redundancy. This effectively improves the efficiency and accuracy of dust monitoring, while reducing the system operation cost, providing more reliable and timely dust monitoring data for the livestock breeding warning platform.

[0017] Preferably, the dynamic cleaning of the PM2.5 sensor in step S1 is specifically as follows:

[0018] Real-time detect the high-frequency sampling period end signal of the PM2.5 sensor and the environmental humidity of the target monitoring area. When the high-frequency sampling period end signal is detected, parse the high-frequency sampling period end signal to generate a high-frequency sampling duration window and a cumulative sampling times threshold;

[0019] If the environmental humidity of the target monitoring area is greater than the preset micro-vibration trigger - environmental humidity threshold, activate the micro piezoelectric actuator to perform the self-cleaning protection operation of the PM2.5 sensor;

[0020] If the environmental humidity of the target monitoring area is less than or equal to the preset micro-vibration trigger - environmental humidity threshold, determine whether the high-frequency sampling duration window is greater than the preset micro-vibration trigger - time window threshold, and whether the cumulative sampling times threshold is greater than the preset micro-vibration trigger - sampling times threshold. If the high-frequency sampling duration window is greater than the preset micro-vibration trigger - time window threshold and the cumulative sampling times threshold is greater than the preset micro-vibration trigger - sampling times threshold, activate the micro piezoelectric actuator to perform the self-cleaning protection operation of the PM2.5 sensor.

[0021] By real-time detecting the high-frequency sampling period end signal and the environmental humidity of the target monitoring area, the present invention can flexibly trigger the self-cleaning operation of the sensor according to environmental conditions and sampling situations. In a high-humidity environment, the cleaning mechanism is directly activated, avoiding the aggravation of sensor surface contamination caused by humidity; in a low-humidity environment, a comprehensive judgment is made by combining the high-frequency sampling duration and the cumulative sampling times, ensuring timely cleaning when the sensor contamination risk is high, while avoiding unnecessary cleaning operations, saving energy and equipment life. This not only improves the long-term stability of the sensor but also ensures the accuracy and reliability of the monitoring data.

[0022] Preferably, the real-time acquisition of the first target herd density distribution map and the first herd individual displacement vector sequence in step S2 is specifically as follows:

[0023] The distribution of target livestock herds in the target monitoring area is detected by using an infrared array sensor to generate a sequence of regional infrared thermal maps;

[0024] Each image in the sequence of regional infrared thermal maps is filtered and eliminated to obtain a sequence of filtered regional infrared thermal maps. The specific formula for the filtering and elimination is as follows:

[0025]

[0026] where I filtered (x, y) is the pixel value at the (x, y) position after filtering, and I raw (x + i, y + j) is the pixel value in the original image that is offset by (i, j) relative to (x, y). σ is the standard deviation of the Gaussian kernel, i is the horizontal offset of the neighborhood pixels centered on (x, y), j is the vertical offset of the neighborhood pixels centered on (x, y), π is the ratio of the circumference of a circle to its diameter, and e is the base of the natural logarithm;

[0027] Based on the sequence of filtered regional infrared thermal maps, the density distribution of the target livestock herds is calculated to obtain the first target livestock herd density distribution map. The specific calculation formula for the density distribution of the target livestock herds is as follows;

[0028]

[0029] where ρ ij is the density of the target livestock herds in the grid cell (i, j), N ij is the number of target animals in the grid cell (i, j), and A grid is the area of a single grid cell;

[0030] Based on the adjacent frame thermal maps in the sequence of filtered regional infrared thermal maps, the individual displacement vectors of the target livestock herds are calculated. The specific formula for calculating the individual displacement vectors of the target livestock herds is as follows;

[0031]

[0032]

[0033] where, is the horizontal coordinate of the i-th target animal at time t, is the vertical coordinate of the i-th target animal at time t, Δx i is the displacement vector in the horizontal direction, and Δy i is the displacement vector in the vertical direction;

[0034] According to the individual displacement vectors of the target livestock herds, a first sequence of individual displacement vectors of the livestock herds is constructed.

[0035] The present invention realizes high-precision real-time acquisition of the density distribution of the target herd and the individual displacement vector sequence by using an infrared array sensor in combination with image processing technology. The infrared thermal map is preprocessed by a Gaussian filtering algorithm, effectively eliminating image noise and improving image quality, thereby ensuring the accuracy and reliability of herd density calculation. By calculating the herd density distribution based on the filtered thermal map sequence and calculating the displacement vectors of herd individuals through adjacent frame thermal maps, the dynamic behavior characteristics of the herd can be accurately captured. This can not only reflect the spatial distribution and movement state of the herd in real time, but also provide accurate data support for subsequent biotic disturbance compensation, enhancing the perception ability of the livestock breeding warning platform to the dynamic changes in the breeding environment.

[0036] Preferably, in step S2, constructing the target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence is specifically as follows:

[0037] Construct a first target herd fluid velocity field according to the first herd individual displacement vector sequence;

[0038] Based on the first target herd density distribution map and the first target herd fluid velocity field, construct a target herd kinematic model through the Navier-Stokes equation, where the control equation of the target herd kinematic model is:

[0039]

[0040] where ρ is the dynamic density field in the first target herd density distribution map, v is the target herd fluid velocity field, p is the population pressure, μ is the viscosity coefficient, F ext is the external driving force, and t is the time.

[0041] The present invention constructs a kinematic model by using the Navier-Stokes equation in fluid mechanics based on the herd density distribution and the individual displacement vector sequence, which can abstract the collective behavior of the herd into a fluid dynamics process, thereby accurately describing the movement state and interaction of the herd in space. This can not only reflect the macroscopic movement characteristics of the herd, such as population density changes and flow trends, but also reveal the dynamic impact of herd behavior on the environment through population pressure and fluid velocity field parameters. By combining herd behavior with environmental factors (such as dust distribution), this model provides a scientific basis for subsequent biotic disturbance compensation.

[0042] Preferably, step S3 includes the following steps:

[0043] Step S31: Real-time collect the second target herd density distribution map and the second herd individual displacement vector sequence;

[0044] Step S32: inputting the second target herd density distribution map and the second herd individual displacement vector sequence into the target herd kinematic model for prediction, thereby obtaining a predicted target herd density gradient vector field and a predicted target herd flow velocity field;

[0045] Step S33: Obtaining a target monitoring area layout diagram and a PM2.5 sensor layout diagram;

[0046] Step S34: performing spatial vector superposition on the predicted target livestock herd density gradient vector field, the predicted target livestock herd flow velocity field, and the target monitoring area layout map to obtain a density gradient-flow velocity distribution map;

[0047] Step S35: dividing the density gradient-flow velocity distribution diagram into disturbance intensity levels according to a preset disturbance intensity threshold standard to obtain a disturbance intensity partition diagram, wherein the disturbance intensity partition diagram includes at least one or more of a high disturbance area and a low disturbance area;

[0048] Step S36: Mapping the PM2.5 sensor layout diagram to the disturbance intensity zoning diagram, and recording the device identifiers of the PM2.5 sensors corresponding to the high disturbance areas in the disturbance intensity zoning diagram as a high disturbance sensor identifier set, and recording the device identifiers of the PM2.5 sensors corresponding to the low disturbance areas in the disturbance intensity zoning diagram as a low disturbance sensor identifier set;

[0049] Step S37: performing differential compensation on the dust vertical distribution gradient data according to the high-disturbance sensor identification set and the low-disturbance sensor identification set to obtain first dust vertical distribution gradient data.

[0050] This invention achieves precise bioturbation compensation for dust monitoring data by introducing predictions from a target herd's kinematic model and superimposing them with spatial vectors. By modeling and predicting real-time herd density and displacement vector sequences, density gradients and flow velocity distributions are generated, which in turn delineate high- and low-disturbance areas, providing a clear zoning basis for PM2.5 sensor compensation. This allows for targeted adjustments to dust monitoring data based on the actual impact of herd behavior on dust distribution, effectively eliminating interference from herd activity and improving the accuracy and reliability of dust concentration monitoring.

[0051] Preferably, step S37 includes the following steps:

[0052] The dust vertical distribution gradient data is divided according to the high-disturbance sensor identification set and the low-disturbance sensor identification set to obtain high-disturbance dust data and low-disturbance dust data;

[0053] Performing a first compensation on the high-disturbance dust data to obtain high-disturbance compensated dust data, wherein performing the first compensation is specifically:

[0054] D comp-H = D raw ·(1 - C H );

[0055]

[0056] Wherein, D comp-H is the high - disturbance compensation dust data, D raw is the high - disturbance dust data, C H is the first compensation coefficient, k is the calibration coefficient, TKE is the turbulent kinetic energy, is the predicted target herd density gradient modulus length;

[0057] Perform a second compensation on the low - disturbance dust data to obtain the low - disturbance compensation dust data, wherein the performing of the second compensation is specifically:

[0058] D comp-L = D raw ·(1 - C L );

[0059] Wherein, D comp-L is the low - disturbance compensation dust data, D raw is the high - disturbance dust data, C L is the second compensation coefficient, specifically, C L = 0.5C H ;

[0060] Record the high - disturbance compensation dust data and the low - disturbance compensation dust data as the first dust vertical distribution gradient data.

[0061] In the present invention, for the high - disturbance area, the turbulent kinetic energy and the herd density gradient modulus length are introduced as key variables in the compensation formula, which can accurately quantify the influence of herd activities on the dust distribution, thereby realizing the precise compensation of the high - disturbance dust data. This fully considers the complexity of herd behavior dynamics and effectively eliminates the problem of over - estimation or under - estimation of dust concentration caused by herd activities. For the low - disturbance area, a relatively small compensation coefficient is adopted to avoid over - compensation and ensure the stability of the data.

[0062] Preferably, step S4 includes the following steps:

[0063] Step S41: Detect the end signal of the feeding event in real - time. When the end signal of the feeding event is detected, obtain the layout map of the target monitoring area, wherein the layout map of the target monitoring area marks the positions of each feeding trough in the target monitoring area;

[0064] Step S42: According to the preset spatial gradient division standard and the positions of the respective feeding troughs, perform dust diffusion zoning on the target monitoring area layout diagram to obtain a dust diffusion - spatial gradient zoning diagram, where the dust diffusion - spatial gradient zoning diagram includes a near - zone of dust diffusion and a far - zone of dust diffusion;

[0065] Step S43: Perform zoning calibration on the first dust vertical distribution gradient data according to the dust diffusion - spatial gradient zoning diagram to obtain the second dust vertical distribution gradient data.

[0066] By introducing a zoning calibration mechanism triggered by the end - of - feeding event signal, the present invention significantly improves the accuracy of dust monitoring data during the critical monitoring window period after feeding. After the feeding ends, according to the positions of the feeding troughs and the preset spatial gradient division standard, the target monitoring area is divided into a near - zone of dust diffusion and a far - zone of dust diffusion. This fully considers the diffusion characteristics of dust over time and space after feeding, making the calibration operation more targeted. By performing differential calibration on the dust data in different regions, it can effectively eliminate the influence of the change in dust concentration gradient caused by the feeding operation on the monitoring data, and avoid the compensation lag problem that occurs in the traditional fixed - cycle calibration strategy after feeding. The finally obtained second dust vertical distribution gradient data is closer to the real environmental conditions, providing high - precision and high - timeliness dust monitoring data for the livestock breeding warning platform.

[0067] Preferably, step S43 includes the following steps:

[0068] Step S431: Extract the recorded timestamp in the end - of - feeding event signal, and obtain the current timestamp. Determine whether the current time is within a preset calibration trigger window according to the recorded timestamp and the current timestamp, where the preset calibration trigger window includes a first calibration trigger window and a second calibration trigger window. The first calibration trigger window is the calibration time range corresponding to the near - zone of dust diffusion, and the second calibration trigger window is the calibration time range corresponding to the far - zone of dust diffusion. The first calibration trigger window is larger than the second calibration trigger window;

[0069] Step S432: If the current time is within the first calibration trigger window, collect near - zone dust diffusion baseline data through a high - resolution dust baseline acquisition instrument, where the measurement accuracy of the high - resolution dust baseline acquisition instrument is not lower than 1 μg / m 3 , perform baseline offset compensation on the dust data belonging to the near - zone of dust diffusion in the first dust vertical distribution gradient data according to the near - zone dust diffusion baseline data to obtain near - zone reference - corrected dust data; and perform baseline correction on the PM2.5 sensor according to the near - zone dust diffusion baseline data;

[0070] Step S433: If it is currently in the second calibration trigger window, interpolate and compensate the dust data belonging to the far dust diffusion area in the first dust vertical distribution gradient data to obtain far area interpolated and compensated dust data;

[0071] Step S434: Record the near area baseline corrected dust data and the far area interpolated and compensated dust data as the second dust vertical distribution gradient data.

[0072] By setting the first calibration trigger window and the second calibration trigger window, the present invention can perform refined calibration on the data in the near and far areas of dust diffusion according to the spatio-temporal characteristics of dust diffusion. In the near area of dust diffusion, a high-resolution dust baseline acquisition instrument is used for baseline offset compensation, ensuring high-precision correction of the near area dust data. At the same time, the sensor is baseline corrected to further improve the long-term stability of the sensor. In the far area of dust diffusion, the interpolation compensation method is adopted, which can effectively make up for the spatio-temporal differences in data and ensure the accuracy of the far area dust data. This calibration method based on time window and regional characteristics not only improves the reliability of dust monitoring data, but also enhances the adaptability and early warning ability of the livestock breeding warning platform to the dynamic changes of the breeding environment.

[0073] Preferably, the present invention also provides a data monitoring and processing system for an Internet of Things-based livestock breeding warning platform, which is used to execute the data monitoring and processing method for the Internet of Things-based livestock breeding warning platform as described above. The data monitoring and processing system for the Internet of Things-based livestock breeding warning platform includes:

[0074] A dust monitoring and cleaning module, which is used to collect dust vertical distribution gradient data in real time through a multi-level PM2.5 sensor array and dynamically clean the PM2.5 sensor; upload the dust vertical distribution gradient data to a preset livestock breeding warning platform;

[0075] A behavior monitoring module, which is used to collect the first target herd density distribution map and the first herd individual displacement vector sequence in real time; construct a target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence;

[0076] A biological disturbance compensation module, which is used to perform biological disturbance compensation on the dust vertical distribution gradient data according to the target herd kinematic model to generate the first dust vertical distribution gradient data;

[0077] A data calibration module, which is used to detect the end signal of the feeding event in real time. When the end signal of the feeding event is detected, perform partition calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

[0078] In the present invention, through the dust monitoring and cleaning module, the system can collect dust vertical distribution gradient data in real time and dynamically clean the PM2.5 sensor, effectively solving the problem of baseline drift caused by the accumulation of particulate matter on the sensor surface, avoiding the compensation lag effect caused by the fixed-cycle calibration strategy, and thus providing more accurate and real-time dust monitoring data for the livestock breeding warning platform. The behavior monitoring module constructs a kinematic model of the target livestock group through the real-time monitoring of the livestock density distribution and individual behaviors, providing a scientific basis for subsequent biotic disturbance compensation and further improving the perception ability of the livestock breeding warning platform to the dynamic changes of the breeding environment. The biotic disturbance compensation module uses the kinematic model to accurately compensate the dust data, eliminating the influence of livestock behavior dynamics interference on the monitoring data and ensuring the reliability and stability of the data received by the livestock breeding warning platform. The data calibration module optimizes the accuracy of the dust data by detecting the end signal of the feeding event in real time and performing zonal calibration, especially the data accuracy during the critical monitoring window period after feeding, providing more accurate data support for the livestock breeding warning platform, thereby enhancing its warning ability and regulation accuracy for the breeding environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Other features, objects, and advantages of the present invention will become more apparent by reading the following detailed description with reference to the accompanying drawings:

[0080] Figure 1 The flowchart of the steps of the data monitoring and processing method of the livestock breeding warning platform based on the Internet of Things according to an embodiment is shown.

[0081] Figure 2 The detailed flowchart of the steps of step S4 according to an embodiment is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0082] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0083] In addition, the drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus their repeated description will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0084] It should be understood that although terms such as "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0085] To achieve the above object, please refer to Figures 1 to 2 , the present invention provides a data monitoring and processing method for an Internet of Things-based livestock breeding warning platform, including the following steps:

[0086] Step S1: Real-time collect dust vertical distribution gradient data through a multi-level PM2.5 sensor array, and dynamically clean the PM2.5 sensor; upload the dust vertical distribution gradient data to a preset livestock breeding warning platform;

[0087] Step S2: Real-time collect the first target herd density distribution map and the first herd individual displacement vector sequence; construct a target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence;

[0088] Step S3: Perform bioturbation compensation on the dust vertical distribution gradient data according to the target herd kinematic model to generate the first dust vertical distribution gradient data;

[0089] Step S4: Real-time detect the end signal of the feeding event. When the end signal of the feeding event is detected, perform partition calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

[0090] In this embodiment, in a livestock breeding house, a multi-level PM2.5 sensor array is first installed. The sensor model is "PM-3000", which is distributed at different heights in the breeding house and is used to collect real-time dust vertical distribution gradient data. These sensors are equipped with micro piezoelectric actuators to dynamically clean the sensor surface and prevent dust accumulation from affecting the measurement accuracy. The dust data collected by the sensors is uploaded to a preset livestock breeding warning platform in real time through a wireless communication module. The platform is built based on a cloud server and has data storage and analysis functions. At the same time, an infrared array sensor with the model "FLIR A310" is used to collect the density distribution map and individual displacement vector sequence of the first target livestock group in the breeding house in real time. The infrared sensor captures the thermal radiation information of the livestock group at a frequency of 10 frames per second to generate a sequence of regional infrared thermal maps. The thermal maps are processed by Gaussian filtering using MATLAB software, and the standard deviation of the filter kernel is set to 2 pixels. Then, an image segmentation algorithm is used to identify the heat source centers of each livestock individual, calculate the number of animals in each grid cell, and obtain the livestock group density distribution map. At the same time, by tracking the heat source centers of each animal, the displacement vectors in the horizontal and vertical directions are calculated to obtain the individual displacement vector sequence of the livestock group. Based on the livestock group density distribution map and individual displacement vector sequence, the ANSYS Fluent software is used to construct a livestock group kinematics model. The density distribution map and the fluid velocity field are used as initial conditions and input, and the Navier-Stokes equation is solved numerically to obtain the density gradient and flow velocity distribution of the livestock group. Then, according to the livestock group kinematics model, the collected dust vertical distribution gradient data is compensated for biological disturbance. The compensated data is recorded as the first dust vertical distribution gradient data. After the feeding event ends, the system detects the feeding event end signal and immediately starts the partition calibration program. According to the preset spatial gradient division standard, the breeding house is divided into a near dust diffusion area and a far area. For the near area, a high-resolution dust baseline collection instrument "DustTrak8530" is used to collect baseline data. The dust data in the near area is calibrated according to the baseline data. For the far area, Kriging interpolation is used to compensate the dust data. Finally, the dust data corrected in the near area and the dust data compensated in the far area are combined to obtain the second dust vertical distribution gradient data.

[0091] Preferably, the sampling frequency adjustment rule of each PM2.5 sensor in the multi-level PM2.5 sensor array in step S1 is specifically as follows:

[0092] Real-time detection of the feeding event signal;

[0093] Specifically, in a livestock breeding house, multiple intelligent sensor nodes are installed. These nodes integrate various sensors and are used to monitor various parameters of the breeding environment. Among them, the tool for detecting the feeding event signal is a photoelectric sensor installed near the feeding system. When the automatic feeding system starts and begins to dispense feed, the flow of the feed will block the optical path of the photoelectric sensor, thereby triggering a signal. For example, assume that a photoelectric sensor with the model "OSM-120" is installed above the feed trough of the feeding system. Its working principle is to judge whether an object passes by detecting the occlusion of light. When the feed flows out from the feeding port and passes through the detection area of the photoelectric sensor, the sensor will output a low-level signal, and the duration is the time when the feed passes through, usually ranging from several seconds to dozens of seconds. This low-level signal is the feeding event signal.

[0094] Based on a preset breeding event trigger signal, determine whether the feeding event signal is a feeding period trigger signal;

[0095] Specifically, after receiving the feeding event signal, a preset time schedule will be called to determine whether the signal is a feeding period trigger signal. The preset time schedule is formulated according to the breeding plan and records the daily feeding time range. For example, assume that the breeding plan stipulates that the daily feeding period is from 6:00 - 6:15 in the morning and from 18:00 - 18:15 in the afternoon. The system will obtain the current time. Assume that the current time is 6:05 in the morning. At this time, a feeding event signal is received. By comparing the current time with the preset feeding period, it is found that 6:05 is within the morning feeding period, so it is determined that this feeding event signal is a feeding period trigger signal. If the current time is 7:00 in the morning and a feeding event signal is also received at this time, and the system finds through comparison that 7:00 is not within the preset feeding period, it will determine that this signal is not a feeding period trigger signal.

[0096] If the feeding event signal is not a feeding period trigger signal, identify the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the standard mode, maintain the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the high-frequency monitoring mode, switch the current sampling frequency mode to the standard mode;

[0097] Specifically, it can be assumed that a multi-level PM2.5 sensor array of model "PM-3000" is installed in the breeding house, and each sensor has two sampling frequency modes: standard mode and high-frequency monitoring mode. The sampling frequency of the standard mode is once per minute, while the sampling frequency of the high-frequency monitoring mode is once every 10 seconds. When the livestock breeding warning platform determines that the feeding event signal is not a feeding period trigger signal, it sends an instruction to the PM2.5 sensor through the communication module to query the current sampling frequency mode of the sensor. For example, if the current sensor is in the high-frequency monitoring mode, the system will send a switching instruction to adjust the sampling frequency of the sensor from once every 10 seconds to the standard mode of once per minute. This switching process is achieved by the microcontroller inside the sensor receiving the instruction and adjusting the parameters of its sampling timer. If the sensor is already in the standard mode, the system will not send any instructions and keep the current sampling frequency mode of the sensor unchanged.

[0098] If the feeding event signal is a feeding period trigger signal, the current sampling frequency mode of the PM2.5 sensor is switched to the high-frequency monitoring mode.

[0099] Specifically, when the livestock breeding warning platform determines that the feeding event signal is a feeding period trigger signal, it can immediately send an instruction to all PM2.5 sensors, requiring the sensors to switch to the high-frequency monitoring mode. Taking the sensor of model "PM-3000" as an example, the instruction sent by the system will be transmitted to the microcontroller of the sensor through the wireless communication module. After receiving the instruction, the microcontroller will adjust the parameters of the sampling timer from once per minute in the standard mode to once every 10 seconds in the high-frequency monitoring mode. For example, at the beginning of the feeding period, assuming that the sensor was originally in the standard mode, after the system detects the feeding event signal and confirms it as a feeding period trigger signal, it will complete the sending and receiving of the instruction within a few milliseconds. After receiving the instruction, the sensor will complete the switching of the sampling frequency within less than 1 second, and thus start to collect PM2.5 data at a frequency of once every 10 seconds.

[0100] Preferably, the dynamic cleaning of the PM2.5 sensor in step S1 is specifically as follows:

[0101] Real-time detect the high-frequency sampling period end signal of the PM2.5 sensor and the environmental humidity of the target monitoring area. When the high-frequency sampling period end signal is detected, analyze the high-frequency sampling period end signal to generate a high-frequency sampling duration window and a cumulative sampling times threshold;

[0102] Specifically, in a livestock breeding shed, each PM2.5 sensor is equipped with a high-frequency sampling period monitoring module and an environmental humidity sensor. Assume that the model of the PM2.5 sensor used is "SensAir-Q2", which has a built-in high-frequency sampling period monitoring function and can perform high-frequency monitoring at a sampling frequency of once every 10 seconds. At the same time, an environmental humidity sensor of model "HTU21D" is installed around the sensor to continuously monitor the environmental humidity of the target monitoring area. When the "SensAir-Q2" sensor completes a high-frequency sampling period (for example, with a duration of 15 minutes), it generates a high-frequency sampling period end signal and sends the signal to the livestock breeding warning platform through the internal communication module. At the same time, the "HTU21D" humidity sensor continuously collects environmental humidity data and sends the humidity value (for example, 45%RH) to the livestock breeding warning platform at a frequency of once a minute. The livestock breeding warning platform receives these two signals simultaneously. After the livestock breeding warning platform receives the high-frequency sampling period end signal of the PM2.5 sensor, it analyzes the signal to obtain relevant information about the high-frequency sampling period. Assume that the high-frequency sampling period end signal contains the sampling start time and end time. By calculating the time difference between these two time points, the high-frequency sampling duration window is determined to be 15 minutes. At the same time, the system calculates the cumulative sampling count threshold to be 90 times (15 minutes × 60 seconds / minute ÷ 10 seconds / sampling) based on the sampling frequency of the sensor (once every 10 seconds) and the duration window.

[0103] If the environmental humidity in the target monitoring area is greater than the preset microseismic trigger - environmental humidity threshold, activate the micro piezoelectric actuator to perform the self-cleaning protection operation of the PM2.5 sensor;

[0104] Specifically, the current environmental humidity value (for example, 45%RH) can be compared with the preset microseismic trigger - environmental humidity threshold (for example, 50%RH). If the current environmental humidity is greater than the preset threshold, the system immediately sends an instruction to the PM2.5 sensor to activate its built-in micro piezoelectric actuator. Assume that the sensor uses a micro piezoelectric actuator of model "PZT-5H". When receiving the instruction, the piezoelectric actuator generates high-frequency vibrations (for example, with a frequency of 100kHz) to remove dust and particulate matter on the sensor surface through physical vibration. This self-cleaning process usually lasts from a few seconds to dozens of seconds, and the specific time can be preset according to the cleaning requirements of the sensor. For example, the system can set the cleaning time to 10 seconds to ensure the cleaning effect of the sensor surface while avoiding damage to the sensor caused by excessive vibration.

[0105] If the environmental humidity in the target monitoring area is less than or equal to the preset microseismic trigger - environmental humidity threshold, then determine whether the high - frequency sampling duration window is greater than the preset microseismic trigger - time window threshold, and whether the cumulative sampling times threshold is greater than the preset microseismic trigger - sampling times threshold. If the high - frequency sampling duration window is greater than the preset microseismic trigger - time window threshold, and the cumulative sampling times threshold is greater than the preset microseismic trigger - sampling times threshold, then activate the micro - piezoelectric actuator to perform the self - cleaning protection operation of the PM2.5 sensor.

[0106] Specifically, if the environmental humidity in the target monitoring area is less than or equal to the preset microseismic trigger - environmental humidity threshold (for example, 45%RH ≤ 50%RH), further determine the high - frequency sampling duration window and the cumulative sampling times threshold. Assume that the preset microseismic trigger - time window threshold is 10 minutes, and the preset microseismic trigger - sampling times threshold is 60 times. Compare the actual high - frequency sampling duration window (15 minutes) with the time window threshold (10 minutes), and at the same time compare the cumulative sampling times threshold (90 times) with the sampling times threshold (60 times). If the high - frequency sampling duration window is greater than the time window threshold (15 minutes > 10 minutes), and the cumulative sampling times threshold is greater than the sampling times threshold (90 times > 60 times), it is determined that the sampling environment of the current sensor is relatively complex, and there is a risk of more particulate matter accumulation. Therefore, an instruction will be sent to activate the micro - piezoelectric actuator to perform the same self - cleaning operation as in the previous steps. For example, send a cleaning instruction to the sensor through the communication module. After the sensor receives the instruction, it starts the "PZT - 5H" micro - piezoelectric actuator for 10 - second high - frequency vibration cleaning to remove the dust and particulate matter on the sensor surface.

[0107] Preferably, in step S2, the real - time acquisition of the first target herd density distribution map and the first herd individual displacement vector sequence is specifically as follows:

[0108] Use an infrared array sensor to detect the target herd distribution in the target monitoring area and generate a sequence of regional infrared thermal maps;

[0109] Perform filtering elimination on each image in the sequence of regional infrared thermal maps to obtain a sequence of filtered regional infrared thermal maps. Among them, the specific formula for the filtering elimination is as follows:

[0110]

[0111] Among them, I filtered (x, y) is the pixel value at the (x, y) position after filtering, I raw(x + i, y + j) is the pixel value in the original image that is offset by (i, j) relative to (x, y), σ is the standard deviation of the Gaussian kernel, i is the horizontal offset of the neighboring pixels centered on (x, y), j is the vertical offset of the neighboring pixels centered on (x, y), π is the ratio of a circle's circumference to its diameter, and e is the base of the natural logarithm;

[0112] Calculate the target herd density distribution based on the filtered regional infrared thermal map sequence to obtain the first target herd density distribution map. The specific calculation formula for the target herd density distribution is as follows;

[0113]

[0114] Among them, ρ ij is the target herd density of the grid cell (i, j), N ij is the number of target animals within the grid cell (i, j), A grid is the area of a single grid cell;

[0115] Calculate the target herd individual displacement vector based on the adjacent frame thermal maps in the filtered regional infrared thermal map sequence. The specific formula for calculating the target herd individual displacement vector is as follows;

[0116]

[0117] Among them, is the horizontal coordinate of the i-th target animal at time t, is the vertical coordinate of the i-th target animal at time t, Δx i is the displacement vector in the horizontal direction, Δy i is the displacement vector in the vertical direction;

[0118] Construct the first herd individual displacement vector sequence according to the target herd individual displacement vector.

[0119] Specifically, an infrared array sensor with the model "FLIR A310" can be installed in the monitoring area. This sensor can capture the thermal radiation information of the target herd at a frequency of 10 frames per second and generate a regional infrared thermal map sequence. Transmit these thermal map sequences to the livestock breeding warning platform and use MATLAB software for image processing. For each image, perform filtering processing using the Gaussian filtering algorithm, where the standard deviation σ of the Gaussian kernel is set to 2 pixels. The specific operation is that for each pixel point (x, y) in the image, calculate its filtered pixel value according to the formula where (i, j) represents the offset of the neighboring pixels centered on (x, y). The filtered thermal map sequence is used to calculate the herd density distribution. Divide the monitoring area into grid cells with a side length of 1 meter, count the number of animals N ij in each grid cell, and according to the formula Calculate the herd density of each grid cell, where A grid is the area of the grid cell. Then, based on adjacent frames in the filtered heat map sequence, by tracking the heat source center of each animal, calculate its displacement vectors Δx i and Δy i , according to the formulas and determine the position of each animal at the next moment. Finally, arrange the displacement vectors of all animals in chronological order to construct the first herd individual displacement vector sequence.

[0120] Preferably, in step S2, constructing the target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence is specifically as follows:

[0121] Construct the first target herd fluid velocity field according to the first herd individual displacement vector sequence;

[0122] Based on the first target herd density distribution map and the first target herd fluid velocity field, construct the target herd kinematic model through the Navier-Stokes equation, where the control equation of the target herd kinematic model is:

[0123]

[0124] where ρ is the dynamic density field in the first target herd density distribution map, v is the target herd fluid velocity field, p is the population pressure, μ is the viscosity coefficient, F ext is the external driving force, and t is the time.

[0125] Specifically, in the livestock breeding environment, to construct the kinematic model of the target herd (such as a flock of sheep), first construct the sheep fluid velocity field according to the sheep individual displacement vector sequence. Use MATLAB software to process the sheep individual displacement vector sequence, and calculate the velocity vector of each sheep by the difference method. For example, the time step is set to 0.5 seconds. For the i-th sheep, its displacement vectors at time t are Δx i (t) and Δy i (t), then its velocity vector can be expressed as Next, interpolate the velocity vectors of all sheep spatially to obtain the distribution of the fluid velocity field within the entire target area. The griddata function in MATLAB can be used for spatial interpolation to convert the discrete individual velocity data into a continuous velocity field distribution map, obtaining the sheep flock fluid velocity field v(x, y, t). Then, based on the sheep flock density distribution map and the sheep flock fluid velocity field, a kinematic model of the sheep flock is constructed using the Navier-Stokes equation. When constructing the model, the computational fluid dynamics (CFD) software ANSYS Fluent is used. First, import the sheep flock density distribution map and the sheep flock fluid velocity field data into ANSYS Fluent. In ANSYS Fluent, set the physical properties of the fluid. For example, the viscosity coefficient can be set to 0.02 Pa·s, which depends on the actual situation of the sheep flock movement and empirical estimation. At the same time, set the external driving force according to the actual situation. For example, factors such as feeding attraction (attraction towards the feeding trough) and obstacle repulsion (thrust away from obstacles) can be considered, and their magnitudes can be adjusted according to the actual scenario, assumed to be 0.1 N / m 3 . Then, use the solver in ANSYS Fluent to solve the Navier-Stokes equation. During the solution process, input the density distribution map and the fluid velocity field as initial conditions, and solve the system of equations numerically to obtain the distribution of the population pressure and the further refinement of the velocity field. The specific governing equations are as follows:

[0126]

[0127] where ρ(x, y, t) is the dynamic density field in the sheep flock density distribution map (unit: heads per square meter), v(x, y, t) is the sheep flock fluid velocity field (unit: m / s), p is the population pressure (positively correlated with the density ρ, unit: Pa), μ is the viscosity coefficient (unit: Pa·s), and F ext is the external driving force (unit: N / m 3 ). After the solution is completed, the detailed results of the sheep flock kinematic model can be obtained, including the density, velocity, and pressure distributions of the sheep flock at different times and positions. These results can be visually displayed in the post-processing module of ANSYS Fluent.

[0128] Preferably, step S3 includes the following steps:

[0129] Step S31: Real-time collect the second target herd density distribution map and the second herd individual displacement vector sequence;

[0130] Specifically, the second target herd density distribution map and the second herd individual displacement vector sequence can be collected by referring to the collection steps of the previous example.

[0131] Step S32: Input the second target herd density distribution map and the second herd individual displacement vector sequence into the target herd kinematic model for prediction to obtain the predicted target herd density gradient vector field and the predicted target herd flow velocity field;

[0132] Specifically, the second target herd density distribution map and the second herd individual displacement vector sequence can be input into the target herd kinematic model. This model is constructed based on the Navier-Stokes equation and solved using ANSYS Fluent software. In ANSYS Fluent, the density distribution map and the fluid velocity field are input as initial conditions, and the equations are solved numerically to obtain the predicted target herd density gradient vector field and the predicted target herd flow velocity field. In specific operations, the time step is set to 0.5 seconds, and the spatial grid resolution is 1 m × 1 m. During the solution process, the viscosity coefficient μ = 0.02 Pa·s and the external driving force F ext = 0.1 N / m 3 . Through the post-processing module of ANSYS Fluent, the prediction results can be visually viewed, including the distribution of density gradient and flow velocity.

[0133] Step S33: Obtain the target monitoring area layout map and the PM2.5 sensor layout map;

[0134] Specifically, the target monitoring area layout map can be obtained. This layout map details each functional area of the breeding house, such as the feed trough, drinking area, and rest area. At the same time, a multi-level PM2.5 sensor array is installed. These layout maps are stored in the livestock breeding warning platform and saved in the form of vector graphics. Using AutoCAD software to open these layout maps, the position of each sensor and the division of the monitoring area can be clearly viewed. For example, the sensor model is "PM-3000", distributed at key positions in the breeding house with a spacing of 10 meters.

[0135] Step S34: Perform spatial vector superposition on the predicted target herd density gradient vector field, the predicted target herd flow velocity field, and the target monitoring area layout map to obtain the density gradient-flow velocity distribution map;

[0136] Specifically, the predicted target herd density gradient vector field and flow velocity field data obtained from ANSYS Fluent software can be imported into GIS (Geographic Information System) software, such as ArcGIS. At the same time, the layout map of the target monitoring area and the layout map of PM2.5 sensors are also imported into ArcGIS. In ArcGIS, a vector overlay operation is performed using spatial analysis tools. The specific steps include: georeferencing the density gradient vector field and flow velocity field data with the layout map of the monitoring area to ensure the spatial consistency of the data; then, through the vector overlay function, the density gradient vector field and flow velocity field are overlaid with the layout map of the monitoring area to generate a density gradient - flow velocity distribution map. This distribution map intuitively shows the distribution of herd density gradient and flow velocity in different monitoring areas.

[0137] Step S35: Classify the density gradient - flow velocity distribution map according to the preset disturbance intensity threshold standard to obtain a disturbance intensity zoning map, where the disturbance intensity zoning map includes at least one or more of high - disturbance areas and low - disturbance areas;

[0138] Specifically, in GIS software (such as ArcGIS), the density gradient - flow velocity distribution map can be classified according to the preset disturbance intensity threshold standard. The preset disturbance intensity threshold standard is set based on empirical values and experimental data. For example, areas with a density gradient greater than 0.5 heads per square meter per meter and a flow velocity greater than 1.0 m / s are classified as high - disturbance areas, and areas with a density gradient less than 0.2 heads per square meter per meter and a flow velocity less than 0.5 m / s are classified as low - disturbance areas. In ArcGIS, the conditional rendering tool is used to classify and render the distribution map according to these threshold standards. The specific operation is to set a conditional expression to mark the areas that meet the high - disturbance conditions as red, the areas that meet the low - disturbance conditions as blue, and other areas as green. The finally generated disturbance intensity zoning map intuitively shows the disturbance intensity levels of different areas.

[0139] Step S36: Map the PM2.5 sensor layout map with the disturbance intensity zoning map, and mark the device identifiers of the PM2.5 sensors corresponding to the high - disturbance areas in the disturbance intensity zoning map as the high - disturbance sensor identifier set, and mark the device identifiers of the PM2.5 sensors corresponding to the low - disturbance areas in the disturbance intensity zoning map as the low - disturbance sensor identifier set;

[0140] Specifically, the PM2.5 sensor layout map and the disturbance intensity zoning map can be imported into GIS software (such as ArcGIS) for mapping operations. In ArcGIS, the location information of the sensors is overlaid with the disturbance intensity zoning map through spatial analysis tools. The specific steps include: spatially matching the location coordinates of the sensors with the disturbance intensity zoning map to determine the disturbance intensity area where each sensor is located. Then, according to the area where the sensor is located, its device identifier is classified. For example, the device identifiers of the sensors located in the high-disturbance area are recorded as the high-disturbance sensor identifier set, and the device identifiers of the sensors located in the low-disturbance area are recorded as the low-disturbance sensor identifier set. These identifier sets can be exported in tabular form and stored in the livestock breeding warning platform.

[0141] Step S37: Differentially compensate the dust vertical distribution gradient data according to the high-disturbance sensor identifier set and the low-disturbance sensor identifier set to obtain the first dust vertical distribution gradient data.

[0142] Specifically, for the detailed implementation process of this embodiment, please refer to the sub-steps of step S37.

[0143] Preferably, step S37 includes the following steps:

[0144] Divide the dust vertical distribution gradient data according to the high-disturbance sensor identifier set and the low-disturbance sensor identifier set to obtain high-disturbance dust data and low-disturbance dust data;

[0145] Perform a first compensation on the high-disturbance dust data to obtain high-disturbance compensated dust data, where the performing of the first compensation is specifically:

[0146] D comp-H =D raw ·(1 - C H );

[0147]

[0148] where D comp-H is the high-disturbance compensated dust data, D raw is the high-disturbance dust data, C H is the first compensation coefficient, k is the calibration coefficient, TKE is the turbulent kinetic energy, is the predicted target herd density gradient modulus length;

[0149] Perform a second compensation on the low-disturbance dust data to obtain low-disturbance compensated dust data, where the performing of the second compensation is specifically:

[0150] D comp-L =D raw ·(1 - C L );

[0151] Among them, D comp-L is the low-disturbance compensated dust data, D raw is the high-disturbance dust data, and C L is the second compensation coefficient. Specifically, C L = 0.5C H ;

[0152] The high-disturbance compensated dust data and the low-disturbance compensated dust data are recorded as the first dust vertical distribution gradient data.

[0153] In this embodiment, it is assumed that the sensor identification set in the high-disturbance area includes sensors numbered 1, 3, and 5, and the sensor identification set in the low-disturbance area includes sensors numbered 2, 4, and 6. The original dust data D raw of these sensors is read from the livestock breeding warning platform and classified according to the area where the sensors are located to obtain high-disturbance dust data and low-disturbance dust data. For the high-disturbance dust data, the first compensation is performed. Assume that the calibration coefficient k = 0.01, the turbulent kinetic energy TKE = 0.5, and the predicted target herd density gradient modulus According to the formula the first compensation coefficient C H = 0.01×0.5×0.3 = 0.0015 is calculated. Then, according to the formula D comp-H = D raw ·(1 - C H ), the high-disturbance dust data is compensated. For example, if the original dust data D raw of a certain high-disturbance sensor = 100 μg / m 3 , then the compensated dust data D comp-H = 100×(1 - 0.0015) = 99.85 μg / m 3 . For the low-disturbance dust data, the second compensation is performed. According to the formula C L = 0.5, the second compensation coefficient C L = 0.5×0.0015 = 0.00075 is calculated. Then, according to the formula D comp-L = D raw ·(1 - C L )

[0154] , the low-disturbance dust data is compensated. For example, if the original dust data D raw of a certain low-disturbance sensor = 80 μg / m 3 , then the compensated dust data D comp-L = 80×(1 - 0.00075) = 79.94 μg / m 3Finally, the high-disturbance compensated dust data and the low-disturbance compensated dust data are merged and denoted as the first dust vertical distribution gradient data. These compensated data will be uploaded to the livestock breeding warning platform.

[0155] Preferably, step S4 includes the following steps:

[0156] Step S41: Detect the end signal of the feeding event in real time. When the end signal of the feeding event is detected, obtain the layout map of the target monitoring area, where the positions of each feeder in the target monitoring area are marked in the layout map of the target monitoring area;

[0157] Specifically, in the livestock breeding house, a photoelectric sensor with the model "OSM-120" can be equipped in the feeding system to detect the start and end of the feeding event. When the feeding system completes the feeding operation, the photoelectric sensor will detect that the feed stops flowing and output a low-level signal with a duration of 1 second as the end signal of the feeding event. The livestock breeding warning platform monitors these signals in real time. Once the end signal of the feeding event is detected, the system will immediately retrieve the layout map of the target monitoring area stored in the database. This layout map is a vector map drawn by AutoCAD software, which details the positions of each feeder in the breeding house, including the number and coordinate information of each feeder. For example, Feeder 1 is located at the coordinates (10, 20) meters, and Feeder 2 is located at the coordinates (30, 40) meters.

[0158] Step S42: Perform dust diffusion zoning on the layout map of the target monitoring area according to the preset spatial gradient division standard and the positions of each feeder to obtain a dust diffusion-spatial gradient zoning map, where the dust diffusion-spatial gradient zoning map includes a dust diffusion near zone and a dust diffusion far zone;

[0159] Specifically, the spatial gradient division standard of dust diffusion can be preset in the livestock breeding warning platform. According to experimental data and experience, after feeding, dust will form a concentration gradient distribution in space, and its diffusion range can be divided into a dust diffusion near zone and a dust diffusion far zone. The dust diffusion near zone is defined as the area within 10 meters from the feeder, and the dust diffusion far zone is defined as the area more than 10 meters from the feeder. Use GIS software (such as ArcGIS) to perform spatial analysis on the layout map of the target monitoring area. First, draw a circular area with a radius of 10 meters centered on each feeder and mark it as the dust diffusion near zone; then, mark the remaining area as the dust diffusion far zone. The finally generated dust diffusion-spatial gradient zoning map is saved in the system in the form of a vector map. For example, the area within 10 meters around Feeder 1 is marked as the near zone, and the area outside it is marked as the far zone.

[0160] Step S43: Perform zoning calibration on the first dust vertical distribution gradient data according to the dust diffusion-spatial gradient zoning map to obtain the second dust vertical distribution gradient data.

[0161] Specifically, for the detailed implementation process of this embodiment, please refer to the sub-steps of step S43.

[0162] Preferably, step S43 includes the following steps:

[0163] Step S431: Extract the recorded timestamp in the feeding event end signal, obtain the current timestamp, and determine whether the current is within a preset calibration trigger window according to the recorded timestamp and the current timestamp. Among them, the preset calibration trigger window includes a first calibration trigger window and a second calibration trigger window. The first calibration trigger window is the calibration time range corresponding to the near area of dust diffusion, and the second calibration trigger window is the calibration time range corresponding to the far area of dust diffusion. The first calibration trigger window is larger than the second calibration trigger window;

[0164] Specifically, in the livestock breeding warning platform of the livestock breeding house, the timestamp of the feeding event end signal can be recorded. Suppose the timestamp of the feeding event end signal is "2025-03-24 10:00:00". After the system detects this signal, it immediately obtains the current timestamp, such as "2025-03-24 10:05:00". The preset calibration trigger window is set according to the dust diffusion characteristics. The first calibration trigger window for the near area of dust diffusion is within 10 minutes after the feeding ends, and the second calibration trigger window for the far area of dust diffusion is within 5 minutes after the feeding ends. The system judges whether the current time is within these two windows by calculating the difference between the current timestamp and the timestamp of the feeding event end signal. In this example, the difference between the current time and the feeding end time is 5 minutes. Therefore, the current is within the second calibration trigger window but does not exceed the first calibration trigger window.

[0165] Step S432: If the current is within the first calibration trigger window, collect the near area dust diffusion baseline data through a high-resolution dust baseline collection instrument. The measurement accuracy of the high-resolution dust baseline collection instrument is not less than 1 μg / m 3 , perform baseline offset compensation on the dust data belonging to the near area of dust diffusion in the first dust vertical distribution gradient data to obtain the near area reference corrected dust data; and perform baseline correction on the PM2.5 sensor according to the near area dust diffusion baseline data;

[0166] Specifically, assume that the current time is within the first calibration trigger window. For example, the current time is "2025-03-24 10:08:00", which is 8 minutes different from the feeding end time. The system will start the high-resolution dust baseline collection instrument installed in the near area of dust diffusion, such as model "DustTrak 8530", whose measurement accuracy is 1 g / m 3。The instrument will collect the baseline data of dust diffusion in the near area. Assume that the collected baseline concentration is 55 g / m 3 。Based on this baseline data, the system performs baseline offset compensation on the dust data in the near area among the first dust vertical distribution gradient data. For example, if the original dust data of a certain sensor in the near area is 60 g / m 3 , the compensated dust data is 60 g / m 3 - 55 g / m 3 = 5 g / m 3 。Meanwhile, the PM2.5 sensor is calibrated according to the baseline data to adjust the zero offset of the sensor.

[0167] Step S433: If the current is in the second calibration trigger window, perform interpolation compensation on the dust data in the far area of dust diffusion among the first dust vertical distribution gradient data to obtain the far area interpolation compensated dust data;

[0168] Specifically, assume that the current time is within the second calibration trigger window. For example, the current time is "2025-03-24 10:04:00", which is 4 minutes away from the end time of feeding. Start the program for interpolating and compensating the dust data in the far area of dust diffusion. Assume that the dust sensors in the far area are relatively sparse, and the Kriging interpolation method is used to process the first dust vertical distribution gradient data. For example, the original dust data of a certain sensor in the far area is 30 g / m 3 , and the compensation value calculated by the Kriging interpolation method at this position is 2 g / m 3 , so the compensated dust data is 30 g / m 3 + 2 g / m 3 = 32 g / m 3 。

[0169] Step S434: Record the near area reference corrected dust data and the far area interpolation compensated dust data as the second dust vertical distribution gradient data.

[0170] Specifically, after completing the calibration of the dust data in the near area and the far area, the near area reference corrected dust data and the far area interpolation compensated dust data can be merged to form the second dust vertical distribution gradient data.

[0171] Preferably, the present invention also provides a data monitoring and processing system based on the Internet of Things livestock breeding warning platform for executing the data monitoring and processing method based on the Internet of Things livestock breeding warning platform as described above. The data monitoring and processing system based on the Internet of Things livestock breeding warning platform includes:

[0172] Dust monitoring and cleaning module, which is used to collect dust vertical distribution gradient data in real time through a multi-level PM2.5 sensor array, and dynamically clean the PM2.5 sensors; upload the dust vertical distribution gradient data to a preset livestock breeding warning platform;

[0173] Behavior monitoring module, which is used to collect the density distribution map of the first target herd and the first individual displacement vector sequence of the herd in real time; construct a kinematic model of the target herd based on the density distribution map of the first target herd and the first individual displacement vector sequence of the herd;

[0174] Biological disturbance compensation module, which is used to perform biological disturbance compensation on the dust vertical distribution gradient data according to the kinematic model of the target herd to generate the first dust vertical distribution gradient data;

[0175] Data calibration module, which is used to detect the end signal of the feeding event in real time, and when the end signal of the feeding event is detected, perform partition calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

[0176] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be included in the present invention.

[0177] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A data monitoring and processing method for an Internet of Things-based livestock farming warning platform, characterized in that, It includes the following steps: Step S1: Collect dust vertical distribution gradient data in real time through a multi-level PM2.5 sensor array, and dynamically clean the PM2.5 sensors; Upload the dust vertical distribution gradient data to a preset livestock breeding warning platform; Step S2: Collect the first target herd density distribution map and the first herd individual displacement vector sequence in real time; construct a target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence; Step S3: Perform biological perturbation compensation on the dust vertical distribution gradient data according to the target herd kinematic model to generate the first dust vertical distribution gradient data; Step S4: Detect the end signal of the feeding event in real time. When the end signal of the feeding event is detected, perform zonal calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

2. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, wherein The sampling frequency adjustment rule of each PM2.5 sensor in the multi-level PM2.5 sensor array described in Step S1 is specifically as follows: Detect the feeding event signal in real time; Based on a preset breeding event trigger signal, determine whether the feeding event signal is a feeding period trigger signal; If the feeding event signal is not a feeding period trigger signal, identify the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the standard mode, maintain the current sampling frequency mode of the PM2.5 sensor. If the current sampling frequency mode is the high-frequency monitoring mode, switch the current sampling frequency mode to the standard mode; If the feeding event signal is a feeding period trigger signal, switch the current sampling frequency mode of the PM2.5 sensor to the high-frequency monitoring mode.

3. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, characterized in that The dynamic cleaning of the PM2.5 sensor described in Step S1 is specifically as follows: Detect the high-frequency sampling period end signal of the PM2.5 sensor and the environmental humidity of the target monitoring area in real time. When the high-frequency sampling period end signal is detected, analyze the high-frequency sampling period end signal to generate a high-frequency sampling duration window and a cumulative sampling times threshold; If the environmental humidity of the target monitoring area is greater than the preset microseismic trigger - environmental humidity threshold, activate the micro piezoelectric actuator to perform the self-cleaning protection operation of the PM2.5 sensor; If the environmental humidity of the target monitoring area is less than or equal to the preset microseismic trigger - environmental humidity threshold, determine whether the high-frequency sampling duration window is greater than the preset microseismic trigger - time window threshold, and whether the cumulative sampling times threshold is greater than the preset microseismic trigger - sampling times threshold. If the high-frequency sampling duration window is greater than the preset microseismic trigger - time window threshold, and the cumulative sampling times threshold is greater than the preset microseismic trigger - sampling times threshold, activate the micro piezoelectric actuator to perform the self-cleaning protection operation of the PM2.5 sensor.

4. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, characterized in that, The real-time collection of the first target herd density distribution map and the first herd individual displacement vector sequence described in Step S2 is specifically as follows: Use an infrared array sensor to detect the distribution of the target herd in the target monitoring area to generate a sequence of regional infrared thermal maps; Perform filtering elimination on each image in the sequence of regional infrared thermal maps to obtain a sequence of filtered regional infrared thermal maps. Among them, the specific formula for the filtering elimination is as follows: Among them, I filtered (x, y) is the pixel value at the (x, y) position after filtering, and I raw (x + i, y + j) is the pixel value in the original image that is offset by (i, j) relative to (x, y). σ is the standard deviation of the Gaussian kernel, i is the horizontal offset of the neighboring pixels centered at (x, y), j is the vertical offset of the neighboring pixels centered at (x, y), π is the ratio of a circle's circumference to its diameter, and e is the base of the natural logarithm; Calculate the density distribution of the target livestock herd based on the infrared thermal map sequence of the filtered area to obtain the first target livestock herd density distribution map. The specific calculation formula for calculating the density distribution of the target livestock herd is as follows; Among them, ρ ij is the target herd density of the grid cell (i, j), N ij A grid is the number of target animals in the grid cell (i, j), and A is the area of a single grid cell; Calculate the displacement vector of the target livestock herd individuals based on the thermal maps of adjacent frames in the infrared thermal map sequence of the filtered area. The specific formula for calculating the displacement vector of the target livestock herd individuals is as follows; Among them, is the horizontal coordinate of the i-th target animal at time t, is the vertical coordinate of the i-th target animal at time t, Δx i is the displacement vector in the horizontal direction, Δy i is the displacement vector in the vertical direction; Construct the first livestock herd individual displacement vector sequence according to the target livestock herd individual displacement vector.

5. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, characterized in that In step S2, the construction of the target livestock herd kinematic model based on the first target livestock herd density distribution map and the first livestock herd individual displacement vector sequence is specifically as follows: Construct the first target livestock herd fluid velocity field according to the first livestock herd individual displacement vector sequence; Construct the target livestock herd kinematic model based on the first target livestock herd density distribution map and the first target livestock herd fluid velocity field through the Navier-Stokes equation. The control equation of the target livestock herd kinematic model is: Among them, ρ is the dynamic density field in the first target herd density distribution map, v is the target herd fluid velocity field, p is the population pressure, μ is the viscosity coefficient, and F ext is the external driving force, and t is the time.

6. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, wherein Step S3 includes the following steps: Step S31: Real-time collect the second target livestock herd density distribution map and the second livestock herd individual displacement vector sequence; Step S32: Input the second target livestock herd density distribution map and the second livestock herd individual displacement vector sequence into the target livestock herd kinematic model for prediction to obtain the predicted target livestock herd density gradient vector field and the predicted target livestock herd flow velocity field; Step S33: Obtain the layout map of the target monitoring area and the layout map of the PM2.5 sensors; Step S34: Perform spatial vector superposition on the predicted target livestock herd density gradient vector field, the predicted target livestock herd flow velocity field and the layout map of the target monitoring area to obtain the density gradient-flow velocity distribution map; Step S35: Divide the density gradient-flow velocity distribution map according to the preset disturbance intensity threshold standard to obtain the disturbance intensity zoning map, where the disturbance intensity zoning map includes at least one or more of the high disturbance area and the low disturbance area; Step S36: Map the layout map of the PM2.5 sensors and the disturbance intensity zoning map, and record the device identifiers of the PM2.5 sensors corresponding to the high disturbance areas in the disturbance intensity zoning map as the high disturbance sensor identifier set, and record the device identifiers of the PM2.5 sensors corresponding to the low disturbance areas in the disturbance intensity zoning map as the low disturbance sensor identifier set; Step S37: Differentially compensate the dust vertical distribution gradient data according to the high disturbance sensor identifier set and the low disturbance sensor identifier set to obtain the first dust vertical distribution gradient data.

7. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 6, characterized in that Step S37 includes the following steps: Divide the dust vertical distribution gradient data according to the high disturbance sensor identifier set and the low disturbance sensor identifier set to obtain the high disturbance dust data and the low disturbance dust data; Perform the first compensation on the high disturbance dust data to obtain the high disturbance compensated dust data. The specific operation of performing the first compensation is: D comp-H = D raw ·(1 - C H ); Among them, D comp-H is the high-disturbance compensation dust data, D raw is the high-disturbance dust data, C H is the first compensation coefficient, k is the calibration coefficient, TKE is the turbulent kinetic energy, is the predicted target herd density gradient modulus length; Perform the second compensation on the low disturbance dust data to obtain the low disturbance compensated dust data. The specific operation of performing the second compensation is: D comp-L = D raw ·(1 - C L ); Among them, D comp-L is the low-disturbance compensated dust data, and D raw is the high-disturbance dust data. C L is the second compensation coefficient. Specifically, C L = 0.5C H ; Record the high disturbance compensated dust data and the low disturbance compensated dust data as the first dust vertical distribution gradient data.

8. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 1, characterized in that Step S4 includes the following steps: Step S41: Real-time detect the end signal of the feeding event. When the end signal of the feeding event is detected, obtain the layout map of the target monitoring area, where the positions of each feeder in the target monitoring area are marked on the layout map of the target monitoring area; Step S42: According to the preset spatial gradient division standard and the positions of each feeder, perform dust diffusion zoning on the layout map of the target monitoring area to obtain a dust diffusion-spatial gradient zoning map, where the dust diffusion-spatial gradient zoning map includes a near dust diffusion area and a far dust diffusion area; Step S43: Perform zoning calibration on the first dust vertical distribution gradient data according to the dust diffusion-spatial gradient zoning map to obtain the second dust vertical distribution gradient data.

9. The data monitoring and processing method of the Internet of Things-based livestock breeding warning platform according to claim 8, characterized in that Step S43 includes the following steps: Step S431: Extract the recorded timestamp in the end signal of the feeding event, and obtain the current timestamp. Determine whether the current time is within a preset calibration trigger window according to the recorded timestamp and the current timestamp, where the preset calibration trigger window includes a first calibration trigger window and a second calibration trigger window. The first calibration trigger window is the calibration time range corresponding to the near dust diffusion area, and the second calibration trigger window is the calibration time range corresponding to the far dust diffusion area. The first calibration trigger window is larger than the second calibration trigger window; Step S432: If currently in the first calibration trigger window, collect near-zone dust diffusion baseline data using a high-resolution dust baseline acquisition instrument, where the measurement accuracy of the high-resolution dust baseline acquisition instrument is not less than 1 μg / m 3 , perform baseline offset compensation on the dust data in the first dust vertical distribution gradient data that belongs to the near-zone of dust diffusion based on the near-zone dust diffusion baseline data to obtain near-zone reference corrected dust data; and perform baseline correction on the PM2.5 sensor based on the near-zone dust diffusion baseline data; Step S433: If the current time is within the second calibration trigger window, perform interpolation compensation on the dust data belonging to the far dust diffusion area in the first dust vertical distribution gradient data to obtain far area interpolation compensation dust data; Step S434: Record the near area reference corrected dust data and the far area interpolation compensation dust data as the second dust vertical distribution gradient data.

10. A data monitoring and processing system for an Internet of Things-based livestock breeding warning platform, characterized in that, A data monitoring and processing system for an Internet of Things-based livestock breeding early warning platform for implementing the data monitoring and processing method according to claim 1, the data monitoring and processing system for the Internet of Things-based livestock breeding early warning platform includes: A dust monitoring and cleaning module, configured to collect dust vertical distribution gradient data in real time through a multi-level PM2.5 sensor array, and dynamically clean the PM2.5 sensors; upload the dust vertical distribution gradient data to a preset livestock breeding early warning platform; A behavior monitoring module, configured to collect a first target herd density distribution map and a first herd individual displacement vector sequence in real time; construct a target herd kinematic model based on the first target herd density distribution map and the first herd individual displacement vector sequence; A biological disturbance compensation module, configured to perform biological disturbance compensation on the dust vertical distribution gradient data according to the target herd kinematic model to generate the first dust vertical distribution gradient data; A data calibration module, configured to real-time detect the end signal of the feeding event, and when the end signal of the feeding event is detected, perform zoning calibration on the first dust vertical distribution gradient data to obtain the second dust vertical distribution gradient data.

Images