Farm management system and method based on a sensing protocol

Abstract

This invention relates to the field of big data processing and discloses a farm management system and method based on a sensor protocol, including a data receiving module, a storage module, and a processing module. The processing module parses the communication protocol header of the original data packet sequence, synchronously acquires metadata reflecting the link transmission quality, and calculates the retransmission mean of the regional sampling sources in combination with a spatial topology mapping table, thereby determining the node-specific degradation characterization quantity after offsetting common-mode interference. The processing module uses a negative exponential mapping rule to generate a real-time confidence weight, and when the weight is lower than a threshold, it introduces the variation characteristics of the reference sampling source to perform interpolation correction on the load. This invention achieves precise decoupling between individual hardware degradation and environmental noise, effectively curbs the pollution of analysis models by implicitly poisoned data, and improves the reliability of big data quality cleaning.

Description

Technical Field

[0001] This invention belongs to the field of big data processing, and in particular relates to a farm management system and method based on a sensor protocol. Background Technology

[0002] Currently, large-scale farming has achieved automation in environmental monitoring, precision feeding, and disease prevention and control through the high-density deployment of IoT sensors and big data analysis. In this system, due to the complex physical environment and severe interference of the sensing layer, the data quality collected by the front-end sensor nodes directly affects the accuracy of the back-end decision-making model. It is the logical cornerstone for realizing the joint analysis of heterogeneous IoT data and the core link in building a reliable IoT system architecture. However, the harsh operating conditions of high corrosion and high humidity in farms often cause gradual physical wear and tear on sensors, rather than instantaneous disconnection. Existing technologies mostly use numerical threshold comparison to identify failed data, which can only filter out samples that exceed the limit. However, during the node degradation period, although the communication link layer frequently experiences retransmission and latency fluctuations, its application layer load value is still within the legal range, showing a hysteretic drift state that is logically legal but distorted in trend.

[0003] Besides the aforementioned limitations of the sensing layer hardware, which is prone to gradual wear and tear in harsh physical environments, software control methods also have shortcomings. For example, Chinese invention patent CN120146407B discloses a method and system for monitoring the livestock production environment based on the Internet of Things. This method uses edge computing and cloud models to regionally monitor and fuse application layer loads, and uses clustering or neural network algorithms to identify abnormal fluctuations in numerical values. Such technologies are highly dependent on the statistical characteristics of the application layer values ​​themselves. The premise is that when the sensor data deteriorates, it is inevitably accompanied by significant numerical deviations or distribution shifts. In the dynamic environment of a livestock farm, nodes are in the early stages of gradual physical wear and tear. The error correction mechanism of the communication protocol stack can still maintain the integrity of the business load logic, causing distortion characteristics to bypass the application layer. Feature cleaning strategies in traditional IoT big data acquisition architectures physically isolate the metadata of the communication protocol stack from the business payload content. The processing center only parses the application layer numerical payload, ignoring the transmission quality characteristics generated by the data stream. This results in a lack of necessary technical feedback mechanisms between the IoT sensing layer and the application layer. This design blind spot, stemming from industry inertia, allows implicitly degraded data containing distortion trends to bypass conventional cleaning mechanisms and continuously be injected into the backend analysis model. This makes it impossible for IoT systems to accurately identify source pollution at the ICT underlying architecture level when processing high-concurrency, multi-source heterogeneous big data streams. The long-term accumulation of distorted features inevitably causes nonlinear deviations in the global decision-making model, fundamentally limiting the accuracy of big data systems in predicting environmental evolution trends.

[0004] Therefore, the technical problem to be solved by this invention is how to use the transmission quality metadata of the heterogeneous sensing protocol of the Internet of Things to perform real-time weight mapping on the application layer payload, and combine the spatial topology characteristics of the IoT edge nodes to eliminate network interference in order to achieve self-healing reconstruction of the data chain. Summary of the Invention

[0005] This invention provides a farm management system based on a sensor protocol, comprising: The data receiving module is used to acquire the sequence of raw data messages reported by heterogeneous sensing units in the observation field; The storage module is used to record the spatial topology image table of the heterogeneous sensing units and preset logical parameters, including preset logical radius, reconstruction threshold and weighting coefficient. The processing module, connecting the data receiving module and the storage module, performs quality cleaning of the original data packet sequence through the following steps: Step S1, parse the communication protocol header of the original data packet sequence to obtain the protocol layer retransmission frequency and packet arrival delay jitter variance of the heterogeneous sensing unit within the observation time window; Step S2, retrieve the spatial topology mapping table and calculate the regional reference attenuation based on the retransmission mean of the reference sampling source within the preset logical radius of the heterogeneous sensing unit; Step S3, calculate the difference between the protocol layer retransmission frequency and the regional reference attenuation to determine the node-specific degradation characterization of the heterogeneous sensing unit; Step S4, generate real-time confidence weights through the negative exponential mapping rule between the node-specific degradation characterization and the packet arrival delay jitter variance; Step S5, if the real-time confidence weights are lower than the reconstruction threshold, introduce the real-time changing gradient of the reference sampling source to correct the payload data of the heterogeneous sensing unit and output a feature vector sequence with quality labels.

[0006] Preferably, when the processing module completes step S2, it adopts the following sub-steps: step S21, locking the logical coordinates of the heterogeneous sensing unit; step S22, matching the set of active sampling sources in the spatial topology mapping table whose distance from the logical coordinates is not greater than the preset logical radius; step S23, calculating the average retransmission frequency of each sampling source in the set of active sampling sources, and outputting the regional reference attenuation.

[0007] Preferably, the value of the real-time confidence weight decreases exponentially with the increase of the node-specific degradation characterization quantity, which is used to limit the tensor input weight of the load data into the back-end analysis model.

[0008] Preferably, when the processing module obtains the protocol layer retransmission frequency, it also parses the physical layer signal-to-noise ratio in the communication protocol header and uses the physical layer signal-to-noise ratio as an adjustment factor for the real-time confidence weight.

[0009] Preferably, when the processing module completes step S5, it adopts the following sub-steps: Step S51: Extract the load reference value of the heterogeneous sensing unit within the preset range of historical confidence level; Step S52: Calculate the real-time variation gradient of the reference sampling source in the current processing cycle; Step S53: Fuse the load reference value and the real-time variation gradient to generate a virtual placeholder load to correct abnormal data in the load data.

[0010] Preferably, the system further includes a gateway preprocessing unit, which is used to filter structurally damaged data frames based on cyclic redundancy check codes before the data receiving module obtains the original data packet sequence.

[0011] Preferably, after completing step S4, the processing module encapsulates the real-time confidence weights in the quality descriptor field of the feature vector sequence to achieve tensor quantization of the original data packet sequence.

[0012] Preferably, the reconstruction threshold is calibrated based on the output frequency of the heterogeneous sensing unit; wherein, for heterogeneous sensing units with an output frequency of not less than 10Hz, the reconstruction threshold is 0.85; and for heterogeneous sensing units with an output frequency of less than 1Hz, the reconstruction threshold is 0.60.

[0013] Preferably, after completing the load data correction, the processing module adjusts the load data according to the real-time confidence weight. The changing trend outputs the hardware reliability label of the heterogeneous sensing unit, and the hardware reliability label is associated with the feature vector sequence for storage.

[0014] A method for managing aquaculture farms based on a sensor protocol includes the following steps: Step 1101: Obtain the original data message sequence reported by the heterogeneous sensing units in the observation field, and parse the communication protocol header of the original data message sequence to extract the protocol layer retransmission frequency and message arrival delay jitter variance of the heterogeneous sensing units within the observation time window. Step 1102: Retrieve the spatial topology mapping table of the heterogeneous sensing unit, match the set of active sampling sources of the heterogeneous sensing unit within the preset logical radius, calculate the retransmission average of each sampling source in the set of active sampling sources, and output the regional reference attenuation amount. Step 1103: Determine the node-specific degradation characterization of the heterogeneous sensing unit by subtracting the regional reference attenuation from the protocol layer retransmission frequency through differential operation. Step 1104: Calculate the real-time confidence weight based on the negative exponential mapping rule between node-specific degradation characteristics and message arrival delay jitter variance. Step 1105: Determine whether the real-time confidence weight is lower than the reconstruction threshold; if it is lower than the reconstruction threshold, calculate the real-time changing gradient of the active sampling source set in the current processing cycle, and use the real-time changing gradient to correct the load data of the heterogeneous sensing unit to generate a feature vector sequence with quality labels.

[0015] Compared with existing technologies, the farm management system based on sensor protocols of this invention has the following advantages: 1. In aquaculture farm management, by breaking down the logical barriers between the IoT communication protocol stack and application layer payload data, an IoT data governance mechanism based on cross-layer communication state perception is constructed to achieve real-time perception of the progressive degradation of sensor hardware. This mechanism is essentially a dynamic weight allocation algorithm executed for IoT spatiotemporal big data. By extracting channel state information from the ICT protocol layer, it provides enhanced input with hardware health dimensions to the backend big data analysis engine. This invention extracts metadata such as retransmission count and latency jitter during message transmission, transforming the error correction behavior of the physical layer and link layer into the evaluation dimension of application layer data weights. This breaks the binary assumption in the traditional architecture that the validity of the value equals the reliability of the data. This mechanism enables the system to identify the latent degradation state in the early stage of sensor probe contamination or RF link moisture, that is, before the payload value exceeds the valid threshold range, by the attenuation of the real-time confidence coefficient. It cuts off the injection of distortion features into the global big data state matrix from the source, ensuring that the data samples obtained by the backend analysis model have physical authenticity.

[0016] 2. By introducing differential calculation rules for the baseline attenuation of the regional network, this invention achieves precise decoupling between global network noise and individual hardware faults. The invention uses the communication status of adjacent nodes within a preset spatial topology as a reference benchmark. Through subtraction logic, it eliminates the common-mode bias caused by gateway-side interference or global congestion, and extracts the specific attenuation amount that characterizes the health of a single node. This mechanism avoids global data misjudgment caused by environmental fluctuations, ensuring that under large-scale, high-density dynamic network conditions, the confidence calculation logic can still accurately lock specific edge nodes in the degradation cycle, thereby improving the stability and specificity of outlier feature identification in the big data preprocessing stage. Attached Figure Description

[0017] Figure 1 This is a flowchart of the sensor data quality cleaning process for the communication protocol metadata perception of the present invention; Figure 2 This is a diagram of the full-link logical architecture and multi-source heterogeneous data governance system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, bottom, transverse, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated.

[0020] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0021] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] A farm management system based on a sensor protocol includes: The data receiving module is used to acquire the sequence of raw data messages reported by heterogeneous sensing units in the observation field; The storage module is used to record the spatial topology image table of the heterogeneous sensing units and preset logical parameters, including preset logical radius, reconstruction threshold and weighting coefficient. The processing module, connecting the data receiving module and the storage module, performs quality cleaning of the original data packet sequence through the following steps: Step S1, parse the communication protocol header of the original data packet sequence to obtain the protocol layer retransmission frequency and packet arrival delay jitter variance of the heterogeneous sensing unit within the observation time window; Step S2, retrieve the spatial topology mapping table and calculate the regional reference attenuation based on the retransmission mean of the reference sampling source within the preset logical radius of the heterogeneous sensing unit; Step S3, calculate the difference between the protocol layer retransmission frequency and the regional reference attenuation to determine the node-specific degradation characterization of the heterogeneous sensing unit; Step S4, generate real-time confidence weights through the negative exponential mapping rule between the node-specific degradation characterization and the packet arrival delay jitter variance; Step S5, if the real-time confidence weights are lower than the reconstruction threshold, introduce the real-time changing gradient of the reference sampling source to correct the payload data of the heterogeneous sensing unit and output a feature vector sequence with quality labels.

[0023] Preferably, when the processing module completes step S2, it adopts the following sub-steps: step S21, locking the logical coordinates of the heterogeneous sensing unit; step S22, matching the set of active sampling sources in the spatial topology mapping table whose distance from the logical coordinates is not greater than the preset logical radius; step S23, calculating the average retransmission frequency of each sampling source in the set of active sampling sources, and outputting the regional reference attenuation.

[0024] Preferably, the value of the real-time confidence weight decreases exponentially with the increase of the node-specific degradation characterization quantity, which is used to limit the tensor input weight of the load data into the back-end analysis model.

[0025] Preferably, when the processing module obtains the protocol layer retransmission frequency, it also parses the physical layer signal-to-noise ratio in the communication protocol header and uses the physical layer signal-to-noise ratio as an adjustment factor for the real-time confidence weight.

[0026] Preferably, when the processing module completes step S5, it adopts the following sub-steps: Step S51: Extract the load reference value of the heterogeneous sensing unit within the preset range of historical confidence level; Step S52: Calculate the real-time variation gradient of the reference sampling source in the current processing cycle; Step S53: Fuse the load reference value and the real-time variation gradient to generate a virtual placeholder load to correct abnormal data in the load data.

[0027] Preferably, the system further includes a gateway preprocessing unit, which is used to filter structurally damaged data frames based on cyclic redundancy check codes before the data receiving module obtains the original data packet sequence.

[0028] Preferably, after completing step S4, the processing module encapsulates the real-time confidence weights in the quality descriptor field of the feature vector sequence to achieve tensor quantization of the original data packet sequence.

[0029] Preferably, the reconstruction threshold is calibrated based on the output frequency of the heterogeneous sensing unit; wherein, for heterogeneous sensing units with an output frequency of not less than 10Hz, the reconstruction threshold is 0.85; and for heterogeneous sensing units with an output frequency of less than 1Hz, the reconstruction threshold is 0.60.

[0030] Preferably, after completing the load data correction, the processing module adjusts the load data according to the real-time confidence weight. The changing trend outputs the hardware reliability label of the heterogeneous sensing unit, and the hardware reliability label is associated with the feature vector sequence for storage.

[0031] A method for managing aquaculture farms based on a sensor protocol includes the following steps: Step 1101: Obtain the original data message sequence reported by the heterogeneous sensing units in the observation field, and parse the communication protocol header of the original data message sequence to extract the protocol layer retransmission frequency and message arrival delay jitter variance of the heterogeneous sensing units within the observation time window. Step 1102: Retrieve the spatial topology mapping table of the heterogeneous sensing unit, match the set of active sampling sources of the heterogeneous sensing unit within the preset logical radius, calculate the retransmission average of each sampling source in the set of active sampling sources, and output the regional reference attenuation amount. Step 1103: Determine the node-specific degradation characterization of the heterogeneous sensing unit by subtracting the regional reference attenuation from the protocol layer retransmission frequency through differential operation. Step 1104: Calculate the real-time confidence weight based on the negative exponential mapping rule between node-specific degradation characteristics and message arrival delay jitter variance. Step 1105: Determine whether the real-time confidence weight is lower than the reconstruction threshold; if it is lower than the reconstruction threshold, calculate the real-time changing gradient of the active sampling source set in the current processing cycle, and use the real-time changing gradient to correct the load data of the heterogeneous sensing unit to generate a feature vector sequence with quality labels.

[0032] Example 1: In a large-scale pig farm environmental data continuous monitoring scenario characterized by high-concentration ammonia corrosion and high-frequency physical collisions with livestock, heterogeneous sensing units are exposed to harsh conditions for extended periods. This leads to progressive degradation of the physical probes and RF antennas, causing frequent retransmissions in the underlying communication link. Specifically, in the complex environment of high-concentration ammonia and high humidity in the farm, corrosive dirt and moisture condensation on the surface of the data acquisition probes alter the dielectric properties of the probe's sensitive material and, through outward permeation from the common substrate, simultaneously change the characteristic impedance of the RF antenna feeder integrated on the same board. This common-mode contamination of the physical surface deteriorates the antenna voltage standing wave ratio and significantly reduces signal radiation efficiency, causing the RF signal to distort before reaching the receiver. Logically, this manifests as frequent retransmissions and latency fluctuations triggered by the media access control layer due to the lack of acknowledgment frames. Thus, a physical mechanism is established between the distortion state of the front-end environmental load and the underlying network link. The synchronous evolution of road quality degradation is addressed by traditional data processing logic, which only compares the upper and lower limits of application layer load values ​​for legality. Since the load values ​​in the early stages of degradation exhibit a lag but still fall within the legal threshold range, the system classifies distorted data as healthy samples and continuously incorporates them into the global big data state matrix. This triggers a shift in the feature extraction of the subsequent agricultural environment regulation model. The data receiving module acquires the original data packet sequence reported by temperature and humidity sensors located in the core area of ​​the breeding unit with a data output frequency of 10Hz. The system is configured with a regional aggregation gateway covering a preset logical radius of 50m. This gateway possesses microsecond-level packet parsing capabilities and concurrent computing power to process all active sampling source packets within this logical radius. The aggregation gateway's network card driver layer is equipped with a link status monitoring component independent of the application layer service flow. It directly captures the physical layer and data link layer data frame confirmation status and timestamp identifiers by calling the underlying network socket interface, extracting the retransmission frequency of the protocol layer corresponding to the original data packet sequence. variance of message arrival delay jitter The preprocessing module adds the extracted link layer features as metadata to the tail of the application layer service payload uploaded by the heterogeneous sensing unit and encapsulates them into an extended structure. The processing module directly reads specific fields in the extended structure to obtain the corresponding parameters. The cross-layer acquisition procedure is executed independently by relying on the gateway's local network driver kernel mechanism, without requiring the front-end heterogeneous sensing unit to have global clock synchronization capability or to upload independent link diagnostic messages.

[0033] The processing module parses the protocol stack header data of the raw data packet sequence and extracts the number of protocol layer retransmissions of the target observation object within a 500ms observation time window. variance of message arrival delay jitter Based on the fact that global network congestion in the farm or radio frequency interference on the gateway side causes a synchronous increase in the number of retransmissions across all nodes, the processing module retrieves the spatial topology mapping table from the storage module, matches the set of 8 active sampling sources within the aforementioned 50m preset logical radius, calculates the average number of retransmissions for each sampling source in the active sampling source set, and outputs the regional network baseline attenuation. The processing module calculates the number of protocol layer retransmissions. Compared with the regional network baseline attenuation The difference is used to determine the node-specific attenuation characterization. This subtraction logic cancels out the common-mode interference factor representing global noise, thus reducing the node-specific attenuation of the characteristic quantity. It directly reflects the hardware reliability of the target observation object itself, and solves the technical problem of decoupling global network fluctuations from single-point equipment failures.

[0034] The processing module uses the negative exponential mapping rule to calculate the real-time confidence coefficient. The specific calculation formula satisfies ,in, The retransmission attenuation factor is determined in advance through on-site gradient occlusion calibration experiments. The time delay jitter attenuation factor is determined in advance through probe accelerated aging calibration experiments, and its dimensions are set to... , A dimensionless real-time confidence coefficient characterizing the reliability of load data. To determine the dimensionless node-specific attenuation characterization quantity, The extracted message arrival delay jitter variance is given by the dimensionless expression. The gateway preprocessing unit includes an RF demodulation module that captures the data frame preamble and synchronously measures the physical layer signal-to-noise ratio (SNR) of the received signal during the physical synchronization stage. Based on a lookup table, the physical layer SNR value is mapped to a normalized adjustment factor ranging from zero to one. The processing module then executes a negative exponential mapping rule to generate the basic real-time confidence weights. Then, the base real-time confidence weights will be... The system performs a scalar multiplication operation with the normalized adjustment factor. After the system extracts and updates the product, the final weight index enters the subsequent reconstruction threshold comparison and judgment stage. The transient electromagnetic interference state of the underlying radio frequency band is directly converted into the numerical penalty term of the application layer tensor input node. The lookup table is established in the early stage of deployment through field multipath fading traversal calibration. Its memory data structure contains discrete signal-to-noise ratio numerical keys and corresponding normalized penalty key values. During system processing, the received signal-to-noise ratio is rounded down. When it is in the safe communication range of greater than or equal to 25 dB, the output adjustment factor is 1. When it falls to the extremely bad range of less than 5 dB, the output adjustment factor is zero. For values ​​in the range of 5 to 25 dB, the corresponding discrete decimal key value is extracted and output strictly according to the linear inverse proportional function of the measured packet loss rate in the calibration stage. This establishes a calculable quantitative link between the physical layer anti-interference capability and the application layer data confidence. The processing module determines the real-time confidence coefficient. Is it below the 0.85 reconstruction threshold set for the 10Hz output frequency? When the target object's radio frequency deflection is caused by probe contamination, will its real-time confidence coefficient be affected? When the load decays to 0.42 and falls below the reconstruction threshold, the processing module blocks the write path of the load data into the backend analysis model. It extracts the load baseline values ​​of the target observation object with a real-time confidence coefficient higher than 0.90 in historical records, and simultaneously calculates the real-time gradient changes of the aforementioned eight active sampling sources in the current processing cycle. The processing module fuses the load baseline values ​​and real-time gradient changes according to the inverse distance weight to generate a virtual placeholder load. The virtual placeholder load is used to correct the original data, and a feature vector sequence with quality labels is output. In the virtual placeholder load generation step, the processing module performs a weighted aggregation operation based on the extracted load baseline values ​​and the real-time gradient changes of each reference sampling source. The calculation formula is set as follows , The last application layer load value after mean filtering within the three consecutive observation time windows before the real-time confidence weight of the target observed object falls below the reconstruction threshold. Represents the first active sampling source in the set. The real-time changing gradient is the load difference between the current processing cycle and the previous processing cycle of each reference sampling source. The representative is the first one calculated based on the physical straight-line distance and the real-time confidence coefficient. The processing module assigns weights to each reference sampling source by merging them. The output value of the aggregation operation is written to the corresponding timestamp storage bit to replace the original abnormal load value. To avoid calculation divergence caused by the inverse distance weight approaching infinity when the physical straight-line distance between an active sampling source and the target observation object approaches zero, the processing module configures an anti-singularity judgment rule before performing the above aggregation operation. Specifically, when the physical straight-line distance of the extracted i-th reference sampling source is less than the preset physical lower limit of the minimum deployment spacing of the sensor, the processing module forcibly locks the physical straight-line distance to the preset lower limit value for subsequent weight calculation. This eliminates the risk of extreme value divergence at the mathematical model level, ensuring the logical consistency and numerical stability of the virtual occupant load aggregation calculation system. Through the dynamic weight scheduling and spatial topology interpolation process with underlying communication status awareness, the system cuts off the path for edge nodes in a progressively deteriorating cycle to inject pollution features into the global data lake, ensuring the continuity of heterogeneous data tensors input to the backend agricultural decision engine in terms of time axis and spatial topology, and establishing a data governance architecture that uses protocol layer metadata to manage application layer business steps.

[0035] Example 2: The physical test platform is configured with a convergence gateway and nine isomorphic temperature and humidity sensor nodes distributed within a 50m logical radius. Each of the nine isomorphic temperature and humidity sensor nodes contains one target observation object and eight reference sampling sources. The data output frequency of each node is fixed at 10Hz. Environmental disturbance sources inject broadband Gaussian white noise with a signal-to-noise ratio of 15dB into the gateway's RF receiver to establish a global network congestion state. A polytetrafluoroethylene (PTFE) choke film with a thickness gradient is coated on the surface of the target observation object's RF antenna to create a physical degradation state for the probe. The size of the observation time window for extracting metadata parameters is controlled by the memory occupancy rate of the convergence gateway and the channel noise floor. When the memory occupancy rate of the convergence gateway reaches 75% and the channel noise floor is greater than a preset noise floor threshold, the observation time window is extended to the upper limit of 500ms. Under the current sudden interference condition, the system sets the observation time window to 500ms. In the initial stage of operation, the choke film thickness is 0μm. Under the action of broadband Gaussian white noise, the number of protocol layer retransmissions between the target observation object and the eight reference sampling sources is... Synchronization jumps to an average of 12 times and message arrival delay jitter variance Reaching 18 The processing module retrieves the spatial topology mapping table and calculates the baseline attenuation of the regional network. The number of retransmissions at the protocol layer was 11.8. Subtract the local network reference attenuation Resulting node-specific decay characterization The value is 0.2. The processing module substitutes the values ​​into the calculation formula to obtain the real-time confidence coefficient. The value is 0.95. The system determines that the data is in a healthy state based on the value of the reconstructed threshold greater than 0.85, excluding the calculation deviation caused by global network fluctuations. As the thickness of the choke film increases to 50μm, the processing module of the control group will import the values ​​containing hysteresis characteristics into the backend database because the temperature and humidity load values ​​collected by the front end of the sensor do not exceed the legal threshold range, causing the output deviation rate of the backend environmental control model to climb to 42.7%.

[0036] Protocol layer retransmission count of the target observation object in the experimental group The number of times the regional network baseline attenuation was increased to 45 during the same period, compared to the eight reference sampling sources. Maintained at 12.5 times, node-specific decay characterization quantity The jump reached 32.5 times, and the processing module calculated the real-time confidence coefficient. The data exhibits a non-linear drop and bottoms out at 0.31. The processing module reconstructs the real-time confidence coefficient based on a threshold below 0.85. A spatial topology interpolation loop was triggered to extract historical real-time confidence coefficients higher than 0.90, and virtual occupant loads were generated by fusing real-time variation gradients from the remaining eight active sampling sources. The root mean square error of the feature vector sequence output by the experimental group compared with the interference-free ideal acquisition sequence was controlled within 3.2%. The real-time confidence coefficient of the out-of-range control group data was set to an observation time window of 50ms. The frequency of oscillations in the range of 0.6 to 0.9 caused frequent triggering of the interpolation correction logic and resulted in a deterioration of the root mean square error of the output sequence to 18.5%. Quantitative test data confirmed that the differential comparison and negative exponential weight mapping calculation rules based on protocol layer metadata have the ability to isolate global network noise and single-point hardware degradation. The calculation rules maintain the numerical consistency of the back-end input feature tensor when the node hardware is in the implicit degradation stage.

[0037] Example 3: In the dynamic feeding scenario of a large-scale pig farm, livestock exhibit regular aggregation and migration, and physical movement alters the local temperature and humidity microclimate distribution. The fixed spatial topology mapping table cannot match the microclimate drift caused by the dynamic distribution of the livestock. When the target observation object experiences hardware degradation and triggers interpolation correction logic, the reference sampling source called by the processing module loses its spatial environmental relevance, resulting in spatial interpolation bias and causing feature extraction offset in subsequent analysis models. The processing module initiates the dynamic calibration process of the spatial topology mapping table to obtain the feeding time axis signal of the farm's feeding system and the livestock density thermal value output by the video monitoring unit. The generation process of the livestock density thermal value is as follows: the image processing engine in the video monitoring unit samples the monitoring screen frame by frame, calls the trained target detection deep neural network to identify the pixel-level bounding box of each live animal, and maps its coordinates onto a pre-divided equidistant virtual grid surface. The processing module counts the total number of bounding box center points falling into each virtual grid and calculates the ratio of this total to the maximum capacity of a single circle. Then, it performs a two-dimensional Gaussian smoothing filter to eliminate boundary noise. The output two-dimensional matrix with spatial gradient smoothing properties is the herd density heat value. This successfully transforms the unstructured image visual stream into a structured tensor feature that can be directly processed by the logic control end. The processing module constructs a dynamic logic radius determination model. When the herd density heat value is greater than a preset congestion threshold, the processing module reduces the preset logic radius to improve the correlation matching degree. When the herd density heat value is lower than the preset congestion threshold, the processing module expands the preset logic radius to increase the number of active sampling sources. Within a specific feeding cycle, when the herd density heat value reaches the preset congestion threshold of 80 heads per circle, the processing module updates the preset logic radius from the initial 50m to 25m based on the dynamic logic radius determination model.

[0038] The processing module re-defines the active sampling source set based on the updated preset logical radius, and extracts the target observation object from the active sampling source set. Physical straight-line distance between active sampling sources and the real-time confidence coefficient of the active sampling source. The processing module uses a product logic of inverse distance ratio and confidence weighting to calculate the fusion allocation weights for specific active sampling sources. The specific calculation formula satisfies ,in, The environmental attenuation constant was obtained in advance through offline microclimate wind tunnel calibration experiments. Assign weights to the dimensionless fusion generated by the calculation. The extracted dimensionless real-time confidence coefficient. To determine the physical straight-line distance, the processing module iterates through all nodes in the active sampling source set and assigns the fusion allocation weight. Combining the fusion allocation weights of each node, it calculates and generates a virtual occupancy payload for the target observation object. The system dynamically shrinks or expands the underlying reference topology boundary using the time axis signal of the fusion aquaculture business actions and herd density characteristics. This, along with the fusion allocation weight calculation formula based on the physical straight-line distance and real-time confidence coefficient, completes the data repair process. This topology evolution and dynamic weighting logic suppresses the error amplification phenomenon of static spatial interpolation under herd movement conditions, maintaining the data foundation numerical consistency of the farm environment big data control system in agricultural business steps. When the system faces extreme conditions such as strong electromagnetic pulse interference superimposed on high-density metal fence radio frequency multipath reflections within the farm, the processing module initiates a dynamic smoothing filtering procedure for the metadata parameters. The processing module uses a moving average window model to adjust the original protocol layer retransmission frequency. To implement noise reduction, the sliding window length is set to... And the step size is ,in, The number of sampling points is determined based on the heartbeat cycle of the sensing unit. For a preset displacement increment, the processing module calculates the arithmetic mean of the sampling points within the window to generate a smooth retransmission frequency. This calculation process filters out non-degraded burst retransmission spikes caused by livestock momentarily blocking the radio frequency antenna, and the processing module simultaneously monitors and smooths the retransmission frequency. The trend of the first derivative change: when the first derivative is greater than the preset degradation growth slope threshold in three consecutive sampling periods, the processing module determines that the current link fluctuation is caused by the irreversible degradation of the hardware entity rather than environmental transient noise. This dynamic smoothing mechanism provides a high signal-to-noise ratio input source for the extraction of node-specific attenuation characteristics, ensuring the stability of the confidence evaluation logic in complex radio frequency interference environments.

[0039] Example 4: When the system faces the pre-deployment calibration conditions before large-scale pig farm deployment, the processing module calibrates environmental parameters by combining the microclimate distribution and radio frequency attenuation characteristics of the sensing unit; the wind tunnel simulation chamber has a built-in controllable ammonia generator, and the test bench fixes the temperature and humidity sensor in a standard state at a preset coordinate node. The signal source injects a reference test message into the antenna end of the temperature and humidity sensor; the control unit increases the ammonia concentration in a gradient to form a physical degradation state of the probe, and the control unit simultaneously adjusts the physical straight-line distance between the temperature and humidity sensor and the receiving gateway. The data acquisition card records the protocol layer retransmission frequency and message arrival delay jitter variance of the traversal concentration and distance matrix. The processor generates a low-level attenuation data tensor that maps the farming conditions based on the protocol layer retransmission frequency and message arrival delay jitter variance.

[0040] The processing module reads discrete values ​​from the underlying attenuation data tensor. It then uses a least-squares algorithm to fit the envelope curve of the protocol layer retransmission frequency as ammonia concentration fluctuates, calculating a dimensionless retransmission frequency attenuation factor. Simultaneously, the module fits the evolution trajectory of the message arrival delay jitter variance, calculating a dimensionless attenuation factor. Delay jitter attenuation factor This processing module combines spatial attenuation slope calculations to generate a dimensionless quantity. Environmental attenuation constant This processing module will use the retransmission frequency attenuation factor. Delay jitter attenuation factor and environmental attenuation constant The system writes the parameters to the storage module and extracts them as the baseline values ​​for determining the confidence level of the target observation object and updating the topological logical radius.

[0041] Example 5: In the scenario of offline calibration and extreme fault-tolerant deployment of global initialization of the underlying sensor network in a large-scale pig farm, the environmental data cleaning logic relies on the objective physical basis of the reconstruction threshold setting. The processing module extracts historical messages covering the complete degradation cycle of the target observation object and the reference temperature and humidity physical true values ​​synchronously collected by a high-precision environmental detector. The processing module calculates the real-time confidence coefficient of the historical message in each observation time window and the corresponding mean square error of temperature and humidity measurement. The processing module constructs a scatter mapping matrix with the real-time confidence coefficient as the independent variable and the mean square error of temperature and humidity measurement as the dependent variable. The processing module combines a local weighted scatter smoothing algorithm to locate the nonlinear inflection point where the derivative of the dependent variable suddenly increases as the independent variable decreases. The system extracts the coordinate values ​​of the independent variable corresponding to the nonlinear inflection point and solidifies them as the reconstruction threshold in the business logic.

[0042] During the online operation cycle, the processing module synchronously extracts the status of reference sampling sources within a preset logical radius. The processing module counts the total number of available reference sampling sources within the preset logical radius whose real-time confidence coefficient is greater than the reconstruction threshold. The spatial topology failure judgment rule set by the system satisfies the inequality. ,in, This is the total number of available reference sampling sources. To pre-define the lower limit threshold based on the convergence boundary of the spatial interpolation algorithm, when the real-time confidence coefficient of the target observation object is lower than the reconstruction threshold and satisfies the inequality, the processing module blocks the triggering command of the spatial topology interpolation loop. The processing module retrieves the historical reference load sequence of the target observation object before the current time node. The processing module uses the autoregressive moving average algorithm to process the historical reference load sequence to calculate and generate the predicted placeholder load in the time dimension. The processing module uses the predicted placeholder load to take over the data filling process. The system uses the time dimension degradation calculation logic to block the divergence path of the interpolation matrix caused by common mode physical damage in extreme areas, and maintains the continuous output state of the underlying load of the aquaculture farm environmental big data control system under the condition of local sensing cluster paralysis.

[0043] When the system executes the spatial topology interpolation loop, the weight allocation for the reference sampling source set follows a cooperative redundancy guarantee procedure. The processing module retrieves the hardware version number and the timestamp of the last online calibration for each reference sampling source from the storage module to construct a hardware reliability evaluation vector. The processing module then calculates the fusion allocation weights. Introducing a reliability compensation operator during the process The calculation rules satisfy ,in, The dimensionless attenuation coefficient is calculated based on the timestamp of the last online calibration. In practice, the specific conversion and calculation process of the dimensionless attenuation coefficient is as follows: The processing module obtains the difference between the current system timestamp and the timestamp of the last online calibration as the running time, and performs logarithmic normalization on the ratio of the preset sensor design life cycle to the running time. The calculated value is directly mapped to the absolute range of zero to one as a reliability compensation operator. This quantification rule accurately converts the fatigue physical properties of the sensor node's natural aging over time into the weight penalty factor in the interpolation model, so that the confidence allocation during data reconstruction has an objective time-domain basis of the hardware service life. When the online running time of an active sampling source exceeds the preset upper limit of the calibration-free period, the processing module automatically reduces the reliability compensation operator index of that node and shifts the weight of data repair to the healthier nodes that have been recently calibrated and have a higher real-time confidence coefficient. This procedure eliminates the interpolation benchmark inaccuracy caused by implicit common-mode drift within the reference sampling source set and establishes the parameter self-healing capability of the big data processing system within the cross-year operating cycle.

[0044] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A farm management system based on a sensor protocol, characterized in that, include: The data receiving module is used to acquire the raw data message sequence reported by heterogeneous sensing units in the observation field; The storage module is used to record the spatial topology image table of the heterogeneous sensing units and preset logical parameters, including preset logical radius, reconstruction threshold and weighting coefficient. The processing module, which connects the data receiving module and the storage module, is used to complete the quality cleaning of the original data message sequence through the following steps: Step S1, parse the communication protocol header of the original data message sequence and obtain the protocol layer retransmission frequency and message arrival delay jitter variance of the heterogeneous sensing unit within the observation time window; Step S2: Retrieve the spatial topology mapping table and calculate the regional reference attenuation based on the average retransmission value of the reference sampling source within the preset logic radius of the heterogeneous sensing unit. Step S3: Calculate the difference between the protocol layer retransmission frequency and the regional reference attenuation to determine the node-specific degradation characterization of the heterogeneous sensing unit; Step S4: Generate real-time confidence weights by using the negative exponential mapping rule between the node-specific degradation characterization and the message arrival delay jitter variance. Step S5: When the real-time confidence weight is lower than the reconstruction threshold, the real-time changing gradient of the reference sampling source is introduced to correct the load data of the heterogeneous sensing unit, and a feature vector sequence with quality label is output.

2. The farm management system based on a sensor protocol according to claim 1, characterized in that, When the processing module completes step S2, it adopts the following sub-steps: Step S21, lock the logical coordinates of the heterogeneous sensing unit; Step S22, match the set of active sampling sources in the spatial topology mapping table whose distance from the logical coordinates is not greater than the preset logical radius; Step S23, calculate the average retransmission frequency of each sampling source in the set of active sampling sources and output the regional reference attenuation.

3. The farm management system based on a sensor protocol according to claim 1, characterized in that, The real-time confidence weight decreases exponentially with the increase of the node-specific degradation characterization, and is used to limit the tensor input weights of the load data entering the back-end analysis model.

4. A farm management system based on a sensor protocol according to claim 1, characterized in that, When obtaining the protocol layer retransmission frequency, the processing module also parses the physical layer signal-to-noise ratio in the communication protocol header and uses the physical layer signal-to-noise ratio as an adjustment factor for the real-time confidence weight.

5. A farm management system based on a sensor protocol according to claim 1, characterized in that, When the processing module completes step S5, it adopts the following sub-steps: Step S51: Extract the load reference value of the heterogeneous sensing unit within the preset range of historical confidence level; Step S52: Calculate the real-time variation gradient of the reference sampling source in the current processing cycle; Step S53: Fuse the load reference value and the real-time variation gradient to generate a virtual placeholder load to correct abnormal data in the load data.

6. A farm management system based on a sensor protocol according to claim 1, characterized in that, The system also includes a gateway preprocessing unit, which filters structurally damaged data frames based on cyclic redundancy check codes before the data receiving module acquires the original data packet sequence.

7. A farm management system based on a sensor protocol according to claim 1, characterized in that, After completing step S4, the processing module encapsulates the real-time confidence weights in the quality descriptor field of the feature vector sequence to achieve tensor quantization representation of the original data packet sequence.

8. A farm management system based on a sensor protocol according to claim 1, characterized in that, The reconstruction threshold is calibrated based on the output frequency of the heterogeneous sensing unit; for heterogeneous sensing units with an output frequency of not less than 10Hz, the reconstruction threshold is 0.85; for heterogeneous sensing units with an output frequency of less than 1Hz, the reconstruction threshold is 0.

60.

9. A farm management system based on a sensor protocol according to claim 1, characterized in that, After completing the load data correction, the processing module outputs the hardware reliability label of the heterogeneous sensing unit according to the changing trend of the real-time confidence weight, and stores the hardware reliability label in association with the feature vector sequence.

10. A method for managing a livestock farm based on a sensor protocol, used to run the livestock farm management system based on a sensor protocol as described in claim 1, characterized in that, Includes the following steps: Step 1101: Obtain the original data message sequence reported by the heterogeneous sensing units in the observation field, and parse the communication protocol header of the original data message sequence to extract the protocol layer retransmission frequency and message arrival delay jitter variance of the heterogeneous sensing units within the observation time window. Step 1102: Retrieve the spatial topology mapping table of the heterogeneous sensing unit, match the set of active sampling sources of the heterogeneous sensing unit within the preset logical radius, calculate the retransmission average of each sampling source in the set of active sampling sources, and output the regional reference attenuation amount. Step 1103: Determine the node-specific degradation characterization of the heterogeneous sensing unit by subtracting the regional reference attenuation from the protocol layer retransmission frequency through differential calculation. Step 1104: Based on the node-specific degradation characterization quantity and the packet arrival delay jitter variance The negative exponential mapping rule is used to calculate the real-time confidence weight; Step 1105: Determine whether the real-time confidence weight is lower than the reconstruction threshold; If the value is below the reconstruction threshold, the real-time gradient of the active sampling source set in the current processing cycle is calculated, and the load data of the heterogeneous sensing unit is corrected using the real-time gradient to generate a feature vector sequence with quality labels.

Images