Beef cattle growth parameter fusion weighing early warning system

Abstract

The application discloses a beef cattle growth parameter fusion weighing early warning system, which comprises individual identification, split load weighing, growth parameter acquisition, quality event acquisition, breeding information interface, re-measurement execution and fusion early warning controller. The system compensates the total load in position-velocity combination according to the consistency of the load center and the limb support polygon, and obtains a reliable observed weight; a drinking water rapid temporary storage state and a feed slow temporary storage state are established, a physiological anchor point weighing event is used to correct state transition parameters and extract net growth weight; further, measurement consistency, mass conservation, individual growth, and the residual fingerprints of the common exposed cattle group and the independent exposed cattle group relative to their respective historical baselines are constructed, candidate abnormal sources are generated, and the re-measurement action with the maximum separability of the expected residual interval is selected for verification. After confirming the abnormal source, the corresponding early warning is output, and the parameter update channel is opened or the affected baseline is frozen according to the abnormal type; when the abnormal source is not confirmed, the abnormal data update is frozen.

Description

Technical Field

[0001] This invention relates to the fields of livestock breeding automation, livestock dynamic measurement and growth status monitoring technology, and more specifically, to a beef cattle growth parameter fusion weighing and early warning system. Background Technology

[0002] In large-scale beef cattle farming, body weight and its stage changes are important bases for evaluating fattening efficiency, adjusting diets, and identifying growth abnormalities. Existing technologies have focused on research in automatic identification, dynamic weighing, and group management. CN110036919A discloses a livestock dynamic weighing device and method, which collects real-time weight through a weighing mechanism, obtains reference weight using empirical mode decomposition, and controls the grouping gate according to a preset weight range; however, the reference weight of this scheme mainly comes from the overall weight sequence, and the effectiveness of the force has not yet been verified by utilizing the load distribution at each bearing position and the geometric relationship of the limb support. CN108151838B discloses a weight measurement method and device, which identifies livestock through electronic tags and obtains weight using a weighbridge, which can improve the efficiency of automatic weighing, but does not establish a joint compensation relationship for the dynamic deviations caused by the speed, bearing position, and trunk movement of beef cattle. CN111796566A discloses a beef cattle information collection and automatic grouping system, which connects electronic identification, intelligent weighing, ultrasonic diagnosis, and grouping execution units for information collection and multi-rule grouping, with a focus on the integration of collection and execution devices.

[0003] Existing methods, even after obtaining weighing values, still struggle to distinguish between short-term reversible mass from feed intake, water intake, and excretion and net tissue growth. They also struggle to determine whether anomalies originate from individual growth, shared areas, feed batches, weighing platforms, or state models. Furthermore, when anomalous data directly participates in historical baseline updates, it can lead to baseline drift and persistent false alarms. Therefore, it is necessary to construct an early warning system that integrates load-support geometry verification, position-velocity compensation, dual-temporary reversible mass separation, physiological anchor point calibration, multi-domain residual attribution, active retesting, and selective parameter updates. Summary of the Invention

[0004] The purpose of this invention is to provide a beef cattle growth parameter fusion weighing early warning system to solve the problems in the prior art where dynamic weighing results are easily affected by support position, passage speed and trunk movement, the observed weight is difficult to separate the reversible mass formed by feeding, drinking and excretion, and the source of abnormal growth is difficult to verify and distinguish.

[0005] The present invention achieves the above objectives through the following technical solution: a beef cattle growth parameter fusion weighing and early warning system is provided, including an individual identification device, a weighing platform, a load detection unit arranged in sections on the weighing platform, a limb landing point acquisition device, a trunk posture and body size acquisition device, a quality event acquisition device, a breeding information interface, a retest execution mechanism, an early warning output device, and a fusion early warning controller connected to the aforementioned devices. The fusion early warning controller will generate an individual multi-source event data package from the target beef cattle’s individual identifier, load sequence within the weighing event recording interval, limb landing point, trunk posture and body size parameters, as well as feed quality inflow, water quality inflow, excretion events and breeding information within the physiological event interval associated with the current weighing event, according to a unified time reference. The load center is determined based on the load sequence, and a support polygon is constructed based on the limb landing points. When the load center is located in the effective inner region of the support polygon, the total load sequence of velocity compensation is determined based on the load center position and the longitudinal displacement of the load center and the torso center of mass. The reliable observed weight is obtained based on the consistency of the compensation load, the load center trajectory, and the torso vertical acceleration. A dual temporary state is established, and the dual temporary state is corrected using physiological anchor weighing events. The reversible mass is separated from the reliable observed weight to obtain the net growth weight. Based on the credible observation weight, structural constraint weight, reversible mass change, individual longitudinal baseline, and the historical baselines of both the jointly exposed and independently exposed cattle herds, a normalized multi-domain conserved residual fingerprint is formed. Candidate anomaly sources are generated, and the retesting action with the highest separability of the expected residual interval is selected. The retesting execution mechanism is controlled to execute the retesting and obtain the retesting residual fingerprint. Based on the retest residual fingerprint, the source of the anomaly is identified and an early warning is issued. The corresponding parameter update channel is opened or the affected baseline is frozen according to the anomaly type. If the source of the anomaly is not identified, the anomaly data update is frozen.

[0006] Furthermore, the individual identification device is located at the entrance side of the weighing platform; the load detection unit outputs load values ​​according to a unified sampling clock, and each load value is associated with the acquisition time and the longitudinal and lateral positions of the load detection unit; The limb landing point acquisition device includes a top-view depth camera located above the weighing platform, and the torso posture and body size acquisition device includes a side-view depth camera located to the side of the weighing platform. The quality event acquisition device includes a weighing feeding trough with an individual identifier, a metering waterer with an individual identifier, an excrement quality detection unit, and a cattle body position tracking unit; The breeding information interface outputs the target beef cattle's barn area identifier, feed batch identifier, environmental control area identifier, and growth stage; the fusion early warning controller writes the weighing and posture data within the weighing event recording interval into the weighing subset, and writes the quality events within the physiological event interval, which ends with the current weighing event and covers the most recent concentrated feeding, concentrated drinking, and subsequent excretion events, into the physiological event subset, and associates the two subsets according to the individual identifier and timestamp.

[0007] Furthermore, the fusion early warning controller uses the load value of each load detection unit as the load weight of the corresponding bearing position to determine the longitudinal position and lateral position of the load center, and forms the load center trajectory according to the acquisition time; it selects at least three landing points that are currently in an effective support state from the left forelimb, right forelimb, right hindlimb and left hindlimb. A supporting polygon is formed by connecting the circumference of the beef cattle's torso, and the boundary of the supporting polygon is offset inward by a preset safe distance to form an effective inward shrinkage area; The preset safety distance is determined based on the sampling error of the limb landing point, the hoof width, and the width of the load-bearing influence of the edge of the weighing platform; sampling times where the load center is outside the effective inner contraction area are not included in the calculation time set of the reliable observation weight.

[0008] Furthermore, the position-velocity dynamic deviation response relationship is calibrated using a standard mass load at different load positions and moving speeds. This calibration is only used to characterize the position-velocity system deviation of the weighing platform. The passing speed is determined by the longitudinal displacement of the load center and the torso center of mass within a preset time window; the torso vertical acceleration is obtained by the second-order difference of the filtered torso vertical position; the fusion early warning controller compensates for the total load according to the corresponding deviation, and excludes hoof impact sampling based on the compensation load fluctuation, the continuity of the load center trajectory and the vertical acceleration-load mapping relationship. When all consistency conditions are met, the median of the weights corresponding to the compensated loads within the effective bearing range is determined as the reliable observed weight.

[0009] Furthermore, the rapid temporary storage state for drinking water records the water mass that has not yet been reduced through distribution or excretion after drinking, and the slow temporary storage state for feed records the feed mass that has not yet been reduced through digestion, rumination or excretion after feeding. When a drinking event is detected, the water inflow is increased to a rapid temporary storage state; when a feeding event is detected, the feed inflow is increased to a slow temporary storage state. When an excretion event is detected, the amount of deduction in excretion quality between the two temporary states is determined based on the time interval between the excretion event and the most recent drinking and feeding events, the current rumination status, and the growth stage. The two temporary storage states are transferred according to the event interval and the state retention parameters corresponding to the growth stage. The parameters of the rapid water storage state are less than those of the slow feed storage state. The transfer results are taken as non-negative values. Event intervals exceeding 4 hours are processed in segments. Background mass loss that is not observed separately is included in the state retention parameters.

[0010] Furthermore, the fusion early warning controller determines the current weighing event as a physiological anchor weighing event when the target beef cattle have reached a stable feeding interval since the most recent concentrated feeding, a stable drinking interval since the most recent concentrated drinking, an excretion event has been detected after the most recent concentrated feeding, the rate of change of the sum of the two temporary states is within a stable range, and a reliable observed weight is obtained for the current weighing event. Based on the reliable observed weight changes between adjacent physiological anchor weighing events, and the mass balance relationship formed by the inflow of feed quality, the inflow of drinking water quality, and the outflow of excretion quality within the corresponding time interval, the state retention parameters of the two temporary states and the excretion quality deduction ratio are corrected according to the mass balance relationship. The net growth weight is obtained by subtracting the corrected rapid water storage mass and slow feed storage mass from the current reliable observed weight.

[0011] Furthermore, the body size-weight mapping relationship is a lookup table established by dividing the body length, body height, chest width and trunk contour volume into partitions. Each partition records the median static weight and the allowable range. The fusion early warning controller interpolates adjacent partitions and corrects them according to the median of the attitude residuals corresponding to the trunk pitch angle and lateral tilt angle to obtain the structural constraint weight. The multi-domain conserved residual fingerprint includes measurement consistency residual, mass conservation residual, individual growth residual, common exposure residual, and independent exposure residual; wherein, the common exposure residual and the independent exposure residual are the median growth residuals of each reference beef cattle in the corresponding herd relative to the herd's own historical baseline, respectively. The deviation of each original residual from the midpoint of the normal range is divided by the corresponding half-width to form a dimensionless normalized residual. The fusion early warning controller records the fingerprint according to the direction, amplitude, duration, and time relationship with the quality event of each normalized residual.

[0012] Furthermore, the co-exposed herd consists of reference beef cattle that share barn areas, feed batches, and environmental control areas with the target beef cattle; The independent exposed herd consists of reference cattle that are matched with the target beef cattle in terms of breed, sex, age, entry weight and growth stage, and do not share the barn area or feed batch. When an individual's growth residual exceeds its allowable range, while the co-exposure residual and independent exposure residual are within their respective allowable ranges, a candidate source of individual growth abnormality is generated. When individual growth residuals and common exposure residuals exceed their respective allowable ranges in the same direction within the same statistical period, while independent exposure residuals remain within the allowable range, a common environmental or feed batch anomaly candidate source is generated; when two herds passing through the same weighing platform both exhibit measurement consistency residuals in the same direction that exceed the allowable range, a weighing platform anomaly candidate source is generated; when mass conservation residuals repeatedly exceed limits in the same direction within a preset time window following the same type of feeding event, drinking event, or excretion event, while measurement consistency residuals remain within the allowable range, a reversible mass state anomaly candidate source is generated.

[0013] Furthermore, the retesting execution mechanism includes a diversion mechanism connected to the outlet of the original weighing platform, a backup weighing channel independent of the original weighing platform, a weighing verification device, and a data acquisition node controller; the weighing verification device includes an unloaded zero-point detection unit and a standard mass loading mechanism. The re-testing actions include re-weighing the target beef cattle at the physiological anchor point on the original weighing platform, cross-weighing in the backup weighing channel, re-weighing the reference beef cattle on the original weighing platform, increasing the collection frequency and the number of reference beef cattle for the two exposed cattle groups, and sequentially performing no-load zero-point verification and standard mass range verification on the original weighing platform. The fusion early warning controller determines the expected residual interval under each retest action in a unified dimensionless normalized residual space; when there are fewer than 20 confirmed abnormal samples, the expected residual interval is generated based on the equipment calibration data and the disturbance results of the state parameters; the separability is determined based on the number of non-overlapping intervals, the minimum boundary distance, the amount of overlap and the residual direction, and the largest one is selected.

[0014] Furthermore, the fusion early warning controller controls the diversion mechanism, the backup weighing channel, the weighing verification device, or the data acquisition node controller to execute the selected retest action, and forms a retest residual fingerprint based on the five types of normalized residuals after the retest; When the collected dimensions of the retested residual fingerprint fall into the corresponding expected residual interval and the directions are consistent, a match is determined. The matching dimensions include at least the target residual and two other types of residuals. When only one candidate anomaly source is matched, the source is confirmed. When there are more than two matching results, the action with the highest separability among the unexecuted retest actions is executed. When all are unmatched, an unconfirmed anomaly warning is output. When a weighing platform anomaly is confirmed, only the position-velocity dynamic deviation response relationship is updated; when a reversible mass state anomaly is confirmed, only the corresponding temporary state parameters are updated; in both cases, the updates of other parameters and the three types of baselines by the abnormal data are frozen. When an individual growth anomaly is confirmed, the longitudinal baseline of the target beef cattle individual is frozen; when a common environment or feed batch anomaly is confirmed, the longitudinal baseline of the affected individuals and the baseline of the commonly exposed herd are frozen; the baseline of the unaffected group continues to be updated, and the update of the affected baseline is resumed after two consecutive normal physiological anchor point cycles.

[0015] The beneficial effects of this invention are as follows: 1. This invention uses a load sequence to determine the load center and verifies the current stress state by the geometric relationship between the load center and the effective inward contraction area of ​​the limb support polygon. When the stress state is valid, combining the load center position and the dynamic deviation of the beef cattle passing through the speed compensation platform can reduce the impact of local edge stepping, incomplete support, changes in passage speed and torso undulation on the weighing results, and improve the stability and authenticity of the reliable observed weight.

[0016] 2. This invention establishes a rapid water storage state and a slow feed storage state, respectively. It uses feeding, drinking, rumination and excretion events to describe short-term reversible weight changes in beef cattle, and uses highly reliable weighing results during physiologically stable periods as physiological anchor point correction state transition parameters. It can separate gastrointestinal contents and body fluid changes from observed weight to obtain net growth weight that better reflects tissue growth.

[0017] 3. This invention constructs measurement consistency residuals, quality conservation residuals, individual growth residuals, common exposure residuals, and independent exposure residuals. The two group residuals respectively reflect the group changes of the corresponding cattle herd relative to its own historical baseline, and can distinguish individual growth abnormalities, barn environment or feed batch abnormalities, weighing platform abnormalities, and reversible quality state abnormalities.

[0018] 4. This invention predicts the residual changes under different retesting actions for multiple candidate anomaly sources, selects the retesting action with the highest separability of the expected residual interval of the candidate anomaly source, and controls the diversion mechanism, the original weighing platform, the backup weighing channel, the reference beef cattle collection node or the weighing verification device to perform the retesting, which can reduce invalid repeated weighing and improve the efficiency of anomaly confirmation.

[0019] 5. This invention opens the corresponding parameter update channel or freezes the affected baseline according to the confirmed source of the anomaly: equipment anomalies and status anomalies only update the corresponding parameters, individual anomalies freeze the longitudinal baseline of the target beef cattle, and common environmental or feed batch anomalies freeze the longitudinal baseline of the affected individuals and the baseline of the common exposed cattle herd, which can prevent abnormal data from contaminating normal parameters and baselines. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the overall structure of the beef cattle growth parameter fusion weighing and early warning system of the present invention; Figure 2 This is a flowchart of the beef cattle growth parameter fusion weighing early warning method of the present invention; Figure 3This is a schematic diagram of the load-support geometric consistency and reliable observation weight reconstruction of the present invention; Figure 4 This is a schematic diagram illustrating the dual temporary storage state, physiological anchor point calibration, and multi-domain conserved residual fingerprint construction of the present invention. Figure 5 This is a schematic diagram illustrating the application scenario of the present invention in a large-scale beef cattle fattening farm. Detailed Implementation

[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and specific embodiments. The following embodiments are used to illustrate the system structure, data processing procedure, parameter determination rules, and anomaly feedback method of the present invention, and do not constitute a limitation on the scope of protection of the present invention.

[0022] In this embodiment, the load is uniformly adopted. Quality is uniformly adopted Length and displacement are uniformly adopted The time is uniformly adopted or The speed is uniformly adopted The sampling frequency is uniformly adopted. The rate of change in quality is uniformly adopted. The average daily net growth rate is uniformly adopted. When the load is converted to mass, the standard gravitational acceleration is taken as... .

[0023] This implementation sets up three formulas (1) to (3), which are used to calculate the load center, determine the reliable observation weight, and update the dual temporary state and extract the net growth weight, respectively. Multi-domain residual attribution, retest action selection, and anomaly source confirmation adopt condition combination and sequential comparison, and do not use continuous weight summation of indicators with different dimensions.

[0024] The upper limit threshold used to limit noise, fluctuations, displacement, or interruptions is taken as the 95th percentile of the corresponding parameter in the normal calibration sample; the lower limit threshold used to confirm payload, spatial coverage, or directional consistency is taken as the 5th percentile of the corresponding parameter in the normal calibration sample; the steady-state interval for feeding and drinking is taken as the 90th percentile of the time required for the corresponding physiological state to enter the steady-state range; the normal range of residuals is taken as the 2.5th to 97.5th percentile of the residual distribution of normal samples. The same threshold is used within the same production batch. The relevant parameters are adjusted according to the selective feedback update rule only after the weighing equipment is recalibrated, a change occurs during the growth stage, or the source of the anomaly is confirmed by retesting.

[0025] Example 1:

[0026] like Figures 2 to 4 As shown, S101, constructing multi-source event data packets for individual beef cattle. A dynamic weighing channel is set up at the junction of the daily drinking water channel and the feeding return channel for beef cattle. A weighing platform with a length of 3.20 m and a width of 0.90 m is set up in the weighing channel. The weighing platform is divided into four longitudinal bearing zones along the direction of the beef cattle's movement. Each longitudinal bearing zone has one left-side load detection unit and one right-side load detection unit along the width of the weighing platform, for a total of eight load detection units.

[0027] Each load detection unit has a rated range of 10 kN, and the overall measurement error does not exceed 0.05% of the rated range. It outputs the sub-load values ​​at 100 Hz according to a unified sampling clock. In the calibration samples, the impact duration of the cattle hoof to the ground is mainly distributed between 0.08 s and 0.20 s. Using a 100 Hz sampling frequency, at least eight load sampling points can be obtained within the shortest impact period.

[0028] A top-view depth camera is installed above the weighing platform, and a side-view depth camera is installed to the side of the weighing platform. The top-view depth camera is used to collect the hoof contours and landing points of the left forelimb, right forelimb, left hindlimb, and right hindlimb; the side-view depth camera is used to collect body length, body height, chest width, trunk contour volume, trunk pitch angle, trunk lateral tilt angle, and trunk vertical position.

[0029] Both depth cameras acquired depth images at 30 Hz. When the cattle were moving at the calibrated maximum normal speed of 1.20 m / s, the longitudinal displacement between adjacent depth frames did not exceed 0.04 m, which met the requirements for continuous correlation of limb landing points and recognition of vertical trunk movement.

[0030] An electronic ear tag reader and an entrance trigger sensor are installed at the entrance of the weighing channel. When the entrance trigger sensor detects that a beef cattle have entered the weighing channel, it activates the electronic ear tag reader to read the individual's identification mark and sends the individual's identification mark, the reading time, and the reading status to the fusion early warning controller.

[0031] The quality event acquisition device includes a weighing feeder, a metering drinker, a rumination status acquisition terminal, and an excretion event detection terminal.

[0032] The mass resolution of the weighing feed trough is 0.05 kg. The initial mass of the feed trough is recorded when the target beef cattle enter the feeding position, and the final mass of the feed trough is recorded when they leave the feeding position. The difference between the initial mass and the final mass is subtracted from the feed replenishment mass during the same period to determine the feed intake inflow.

[0033] The volumetric flow rate resolution of the metering waterer is 0.01 L. Based on the cumulative water supply volume and water temperature-corresponding density during the target beef cattle's drinking period, the cumulative water supply volume is converted into drinking water mass inflow.

[0034] The rumination status acquisition terminal determines the start time, end time, and duration of rumination based on periodic neck movements. When the periodic neck movements are sustained for at least 60 seconds, the cattle are confirmed to be in a rumination state. This 60 seconds is taken as the 95th percentile of the duration of short-term chewing (non-rumination) in normal samples.

[0035] The excretion event detection terminal confirms an excretion event based on three conditions: tail root elevation or back posture change, the target cattle being located above the waste collection area, and an increase in the mass of the waste collection area. It outputs the excretion time and mass only when all three conditions are met simultaneously. If the duration of the mass increase in the waste collection area is less than 0.20 s, it is judged as structural vibration and not considered an excretion event.

[0036] The fusion early warning controller provides a unified time reference to the load detection unit, depth camera, and mass event acquisition device. Through one hundred synchronous trigger tests, the maximum link delay of each acquisition node relative to the fusion early warning controller was measured to be 0.018 s; therefore, the allowable time alignment error is set to 0.02 s. Data points with a time difference exceeding 0.02 s are not considered as joint observation data at the same moment.

[0037] When any foreleg of the target beef cattle enters the weighing platform, the integrated early warning controller establishes a weighing event recording interval; 0.50 s after all four limbs of the target beef cattle have left the weighing platform, the weighing event recording interval ends. The 0.50 s is taken as the 95th percentile of the time required for the structural vibration of the weighing platform to recover to the no-load fluctuation range.

[0038] In addition to the weighing event recording interval, the fusion early warning controller also establishes a physiological event association interval with the end time of the current weighing event as the endpoint. The physiological event association interval traces back to the earlier of the most recent concentrated feeding event and the most recent concentrated drinking event, and covers excretion events associated with the target beef cattle after the aforementioned events; when no new quality events occur between adjacent weighing events, the last valid state record of the previous physiological event association interval is used.

[0039] The fusion early warning controller writes the sub-load sequences, limb landing points, trunk postures and body size parameters with the same individual identifier within the weighing event recording interval into the weighing subset, and writes the feed quality inflow, drinking water quality inflow, excretion events and aquaculture information within the physiological event association interval into the physiological event subset, and associates the two subsets according to the individual identifier and timestamp to form an individual multi-source event data packet.

[0040] S102. Filtering valid weighing data based on load-support geometric consistency. Before the system was put into operation, the eight load detection units were calibrated under no-load and under single-shoe load conditions.

[0041] During the no-load calibration, the weighing platform was kept in a no-load state for 30 seconds, and the no-load mean and no-load standard deviation of each load detection unit were calculated. During the single-hoof load calibration, fifty beef cattle with a weight range of 300 kg to 700 kg were selected, and each beef cattle passed through the weighing platform ten times. The single-hoof support range was confirmed based on the depth image, and the sub-load values ​​of the corresponding load detection units were extracted.

[0042] The effective load threshold for each load detection unit is the larger of the following two load values: the load value corresponding to the no-load mean of the load detection unit plus six times the no-load standard deviation, and the 5th percentile of the load values ​​of all confirmed single-shoe support samples.

[0043] In this embodiment, the upper limit of the no-load fluctuation of each load detection unit is between 28 N and 41 N. The 5th percentile of the single-shoe support sample load value is 276 N. Therefore, the effective load threshold of the eight load detection units is uniformly set to 280 N.

[0044] When the load value output by the load detection unit is higher than 280 N for three consecutive sampling moments, the corresponding load detection unit is marked as having an effective load-bearing state; when the load value is lower than 280 N for three consecutive sampling moments, the corresponding effective load-bearing state ends. The three load sampling moments correspond to 0.03 s, which is longer than the duration of system electrical pulse interference but shorter than the duration of normal hoof impact.

[0045] A top-down depth camera detects the hoof profile. When the vertical distance between the lowest point of the hoof and the bearing surface of the weighing platform does not exceed 0.015 m, and this state is maintained for three consecutive depth frames, the corresponding hoof position is determined as the effective support landing point.

[0046] The 0.015 m is taken as the 95th percentile of the vertical distance error between the manually marked hoof landing point and the depth camera recognition result; three depth frames correspond to 0.10 s, which is used to exclude misjudgments caused by missing depth data in a single frame and hoof swaying.

[0047] When there are at least three effective support landing points, the fusion early warning controller connects the current effective support landing points in the circumferential order of the beef cattle's torso to form a support polygon.

[0048] For any sampling time t, the fusion early warning controller determines the load center according to formula (1):

[0049] In the formula, The time of load sampling is expressed in units of 1 / 2000. ; Sampling time A set of indexes for load detection units in an effective load-bearing state; For index set The Middle Each load detection unit at the sampling time Output load value, in units of ; and The first The longitudinal and transverse coordinates of the center position of each load detection unit, in units of 1. ; and Sampling time The corresponding longitudinal and transverse coordinates of the load center, in units of 1. .

[0050] In formula (1) It is used to represent the actual physical contribution of different load-bearing locations to the load center and is not a weight set artificially during the anomaly evaluation process.

[0051] The fusion early warning controller offsets each boundary of the supporting polygon by a preset safe distance inward, forming an effective recessed area. Five hundred normal passage records of fifty beef cattle were selected, and the positional error between manually marked landing points and depth camera-identified landing points was statistically analyzed. The load center offset caused by the edge bearing of the weighing platform was detected using a standard mass block. The sum of these two types of errors was then calculated. The percentile is determined as the preset safety distance. In this embodiment, the preset safety distance is... .

[0052] When the load center is located outside the effective inner contraction area, there are fewer than three effective support landing points, or the movement direction of the four limb landing points is inconsistent with the trajectory direction of the load center, the corresponding sampling time will not be included in the set of reliable observation weight calculation times.

[0053] S103, Implement position-velocity compensation and obtain reliable observation weight. The position-velocity dynamic deviation response relationship of the weighing platform was established using a standard mass calibration trolley. The calibration masses were set sequentially to 300 kg, 400 kg, 500 kg, 600 kg, and 700 kg, covering the normal weight range of the beef cattle to be tested.

[0054] The calibrated trolley's moving speed was set to 0.20 m / s, 0.40 m / s, 0.60 m / s, 0.80 m / s, 1.00 m / s, and 1.20 m / s, respectively, covering the 2.5th to 97.5th percentiles of the normal beef cattle passage speed distribution.

[0055] Each combination of calibrated mass and moving speed was repeated ten times along the center path of the weighing platform, a path offset 0.18 m to the left, and a path offset 0.18 m to the right. The 0.18 m was taken as the 95th percentile of the lateral offset of the load center in the normal passage sample.

[0056] The integrated early warning controller records the total load deviation corresponding to different longitudinal position intervals, lateral position intervals, and movement speed intervals of the load center, forming a position-velocity dynamic deviation calibration table. During actual weighing, adjacent calibration nodes are matched based on the current load center position and the speed at which the cattle pass, and linear interpolation is used to determine the dynamic deviation compensation value.

[0057] The speed of beef cattle is determined by a dual-channel cross-determination: first, the first speed is obtained based on the change in the longitudinal position of the load center within 0.20 seconds, and then the second speed is obtained based on the longitudinal displacement of the torso centroid identified by the side-view depth camera; when the difference between the two exceeds the 95th percentile of the absolute value of the speed difference of normal passing samples, the corresponding time period is not included in the set of valid sampling times.

[0058] The standard mass calibration trolley is only used to establish the systematic deviation of the weighing platform structure under different load positions and moving speeds, and is not used to simulate the impact of beef cattle hooves touching the ground. Hoof impact is identified based on the mapping relationship between trunk vertical acceleration and the rate of change of compensation load established from actual beef cattle calibration samples; trunk vertical acceleration is obtained by performing second-order difference on the trunk vertical position after five-point moving average.

[0059] The weight of a reliable observation is determined according to formula (2):

[0060] In the formula, The reliable observed weight corresponding to the current weighing event, in units of ; This is the set of effective sampling times that simultaneously satisfy both the load-support geometry and motion consistency conditions. Sampling time A set of indexes for load detection units in an effective load-bearing state; For the first Each payload detection unit at the sampling time Output load value, in units of ; and These are the longitudinal and transverse coordinates of the load center, respectively, in units of [missing information]. ; Sampling time The corresponding speed of beef cattle passing through, in units of ; The load compensation value is determined from the position-velocity dynamic deviation calibration table based on the load center position and the passing speed, in units of... ; Let the standard gravitational acceleration be taken as... ; Represents the set of valid sampling times The median value is calculated for each compensated quality value.

[0061] Formula (2) first deducts the position-velocity dynamic deviation from the total load, then converts the compensated load into mass, and finally uses the median of the mass value corresponding to the effective sampling time as the reliable observation weight, without setting artificial weights for different sampling times.

[0062] Enter the set of valid sampling times The sampled data must also meet the following motion consistency conditions: the compensation load is continuous The coefficient of variation within the effective bearing range shall not exceed The load center is in the adjacent The displacement within does not exceed The duration of the interruption in the load center trajectory shall not exceed The sampling proportion of the compensation load change rate falling within the 95% prediction range corresponding to the current torso vertical acceleration range is not less than... .

[0063] in, , and Take the statistic corresponding to the normal passage sample respectively percentile, The sampling proportion of actual beef cattle calibration samples whose rate of change of compensation load falls within the corresponding vertical acceleration prediction interval is taken as the first... Percentile.

[0064] When all motion consistency conditions are met and the number of valid sampling times is not less than fifty, output the reliable observation weight. Fifty load sampling times correspond to It can cover at least one stable support interval. When the number of valid sampling moments is less than fifty, the current weighing event is marked as a low-confidence weighing event, and the weighing result is prohibited from updating the individual longitudinal baseline, the baseline of the co-exposed cattle herd, and the baseline of the independently exposed cattle herd.

[0065] S104. Establish a dual temporary reversible quality state and extract net growth weight. The integrated early warning controller establishes a rapid water storage state and a slow feed storage state for each beef cattle.

[0066] The rapid water storage status is used to record the water mass that has not yet been reduced through distribution or excretion after drinking; the slow feed storage status is used to record the feed mass that has not yet been reduced through digestion, rumination or excretion after feeding.

[0067] To avoid mixing up mass event numbers and weighing event numbers, the serial number of discrete mass events is denoted as k, and the serial number of reliable weighing events is denoted as j. After the k-th mass event occurs, the dual temporary state is updated discretely once; when the j-th reliable weighing event occurs, the dual temporary state is converted to the corresponding weighing time tj, and the reversible mass is separated from the reliable observed weight.

[0068] The dual temporary storage status and net growth weight are determined according to formula (3):

[0069] In the formula, The discrete sequence number of the quality event; and respectively handle the first The rapid temporary storage quality of drinking water and the slow temporary storage quality of feed prior to each quality incident are shown in units of [missing information]. ; and respectively handle the first The rapid temporary storage quality of drinking water and the slow temporary storage quality of feed after each quality incident, in units of... .

[0070] For the first The first quality incident and the first The time interval between quality events, in units of ; For the first The growth stage identifier of the target beef cattle when a quality event occurs.

[0071] To quickly store drinking water in a temporary state at time intervals and growth stage The state retains parameters; For the slow temporary storage of feed at time intervals and growth stage The state-preserving parameters are as follows; both are dimensionless numbers with a range of values ​​of 1. to .

[0072] For the first The drinking water quality inflow corresponding to each quality event, in units of When the first When the quality incident is not a drinking water incident Pick . For the first Feed intake quality inflow corresponding to each quality event, in units of When the first When a quality incident is not a feeding incident. Pick .

[0073] For the first The discharge mass corresponding to each quality event, in units of When the first When a quality incident is not a discharge incident. Pick .

[0074] and The first The percentage reduction in excretion quality corresponding to each excretion event from the rapid temporary storage state of drinking water and the slow temporary storage state of feed, both of which are dimensionless; when Greater than hour, ;when equal At that time, the two deduction ratios do not participate in the status update result.

[0075] The discrete sequence number of the reliable weighing event; For the first The weighing time of a reliable weighing event, in units of ; For the first The trusted observed weight corresponding to each trusted weighing event, in units of ; and The transition from the dual temporary storage state to the weighing time is as follows. The following are the rapid temporary storage quality of drinking water and the slow temporary storage quality of feed, in units of... ; For the first Net growth weight corresponding to each reliable weighing event, in units of .

[0076] At the weighing time When the time coincides with a certain quality event, the state value after processing that quality event is used as... and At the time of weighing When located between two adjacent quality events, the distance will be... Recently and no later than The state value is transferred to the state retention parameters corresponding to the event interval and growth stage. ,get and .

[0077] Under the same event interval and the same growth stage, the state retention parameter of the rapid temporary storage state of drinking water is smaller than that of the slow temporary storage state of feed, reflecting that drinking water quality has a faster in vivo distribution and excretion rate relative to feed quality.

[0078] For each growth stage, the quality event interval is divided into... to When the event interval exceeds and not exceeding Greater than and not exceeding Greater than and not exceeding and greater than At that time, the system divides the system into continuous sub-intervals with a maximum of 4 hours, and performs state transitions sequentially for each sub-interval.

[0079] The candidate state retention parameters for the rapid temporary storage state of drinking water are in to Within the scope according to The interval selection; the candidate state retention parameters for the slow temporary storage state of feed are in to Within the scope according to The interval selection.

[0080] For each candidate parameter combination, the mass conservation residuals of the six most recent adjacent physiological anchor weighing events are calculated. The parameter group with the smallest median absolute value of the mass conservation residuals is selected as the state preservation parameter for the corresponding event interval and growth stage.

[0081] When the excretion event is no more than the most recent drinking event And more than [time period] since the most recent mass feeding incident At that time, Set as ,Will Set as .

[0082] When the excretion event is no more than the most recent mass feeding event And more than [time period] since the most recent drinking water incident At that time, Set as ,Will Set as .

[0083] When an excretion event occurs simultaneously during both the drinking water and feeding periods, or simultaneously exceeds both periods, and All set to .

[0084] The aforementioned deduction ratios are determined based on the median deduction ratio of excrement quality relative to the temporary storage quality of drinking water and feed in manually confirmed excrement source samples. If the mass conservation residuals of six consecutive physiological anchor weighing events all change in the same direction, then... To adjust the step size, modify the corresponding deduction ratio, while maintaining the sum of the two deduction ratios at [value missing]. .

[0085] Each time after updating according to formula (3), if the calculated result of any temporary mass is less than 0 kg, the temporary mass is set to 0 kg, and the unfinished excretion deduction is transferred to another temporary state; if neither temporary state is sufficient to bear the deduction, the remaining part is recorded as unallocated excretion loss, and negative values ​​are not allowed for the temporary mass. The state retention parameter is calibrated by physiological anchor points to simultaneously calculate the respiration, evaporation, and trace background mass loss that is not observed separately.

[0086] Utilizing physiological anchor points to calibrate dual temporary states The current reliable weighing event will be determined as the physiological anchor weighing event when the target beef cattle simultaneously meet the following conditions: the time since the most recent concentrated watering has reached [a certain threshold]. The time since the last mass feeding has reached At least one excretion event has been confirmed since the most recent concentrated feeding; the rate of change of the sum of the rapid temporary storage mass of drinking water and the slow temporary storage mass of feed is continuous. No more than The current weighing event obtains the reliable observed weight according to formula (2).

[0087] Among them, the stable interval of drinking water The time required for the rate of change of drinking water in a rapid temporary storage state to enter a stable range. Percentile; steady-state feeding interval The time required for the rate of change of feed in slow temporary storage to enter a stable range. Percentile; Upper limit of stable rate of change Take the th absolute value of the reversible mass change rate during the normal physiological steady period Percentile; consecutive Used to eliminate short-term fluctuations in physiological state.

[0088] The integrated early warning controller calls the weighing events of the six most recent adjacent physiological anchor points, and corrects the status retention parameters and excretion quality deduction ratio based on the reliable observed weight changes, feed quality inflow, drinking quality inflow, excretion quality outflow and net growth corresponding to the individual longitudinal baseline between adjacent events.

[0089] Each calibration adjusts only one type of parameter. When adjusting the rapid temporary storage parameters for drinking water, freeze the slow temporary storage parameters for feed and the excrement quality deduction ratio; when adjusting the slow temporary storage parameters for feed, freeze the rapid temporary storage parameters for drinking water and the excrement quality deduction ratio; when adjusting the excrement quality deduction ratio, freeze both state retention parameters.

[0090] The adjusted parameters are written into the formal parameter table only if the absolute value of the mass conservation residual decreases in the next physiological anchor weighing event; if the absolute value of the mass conservation residual does not decrease, the parameters before adjustment are restored.

[0091] The reliable observed weight of the first physiological anchor weighing event within a specific weighing cycle. for The corresponding drinking water rapid temporary storage quality for Feed slow storage quality for The net growth weight is obtained according to formula (3). for .

[0092] Fourteen days later, the reliable observed weight of the second physiological anchor weighing event was... The corresponding drinking water rapid temporary storage quality is The slow-storage quality of feed is Net growth weight is .

[0093] Net growth weight increase between two adjacent physiological anchor weighing events The corresponding average daily net growth is .

[0094] S106. Form multi-domain conserved residual fingerprints and generate candidate anomaly sources. The integrated early warning controller establishes a body size-weight correspondence based on calibration samples obtained from verified static weighing equipment. The calibration samples are organized into datasets according to variety, sex, and growth stage, with each dataset containing at least two hundred high-confidence samples.

[0095] Each calibration sample includes body length, body height, chest width, trunk contour volume, trunk pitch angle, trunk lateral tilt angle, and static weighing value. Body length, body height, chest width, and trunk contour volume are divided into corresponding intervals. A lookup table is established between the body size parameter combinations and the static weighing intervals. The static weighing intervals are then corrected based on the trunk pitch angle and trunk lateral tilt angle to obtain the structural constraint weight.

[0096] Specifically, the partition widths for body length, body height, chest width, and trunk outline volume are set to 0.05m, 0.05m, 0.02m, and 0.02m, respectively. 3 The trunk pitch and lateral tilt angles were divided into posture intervals of 2°. Each body size interval contained at least 20 static weighing calibration samples, and the median static weight and the allowable interval formed by the 2.5th percentile to the 97.5th percentile were recorded. When the target beef cattle were located between adjacent intervals, multilinear interpolation was used, and the posture correction was measured as the median of the static weighing residuals within the corresponding posture interval. Intervals with insufficient samples were only merged with adjacent intervals of the same breed, sex, and growth stage.

[0097] The integrated early warning controller determines the measurement consistency residual, mass conservation residual, individual growth residual, common exposure residual, and independent exposure residual, respectively.

[0098] Measurement consistency residuals characterize the deviation of the credible observed weight from the median of the structural constraint weight range; mass conservation residuals characterize the remaining deviation of the credible observed weight change between adjacent weighing events after deducting the reversible mass change and the net growth of the individual baseline during the same period; individual growth residuals characterize the deviation of the current net growth weight of the target beef cattle from the predicted value of the individual longitudinal baseline; co-exposure residuals are the median of the growth residuals of each reference beef cattle in a co-exposed herd relative to the herd's own historical baseline; independent exposure residuals are the median of the growth residuals of each reference beef cattle in an independently exposed herd relative to the herd's own historical baseline.

[0099] Individual longitudinal baselines were established based on the eight most recent physiological anchor weighing events that confirmed normality for the target beef cattle. The co-exposed herd consisted of reference beef cattle that shared barn areas, feed batches, and environmental control zones with the target beef cattle.

[0100] Independently exposed herds consist of cattle of the same breed and sex, with an age difference not exceeding [number missing]. The difference in entry weight should not exceed the target entry weight of the beef cattle. The reference beef cattle composition consists of cattle that are at the same growth stage but do not share the same barn area or feed batch as the target beef cattle.

[0101] The number of reference cattle used to form the baseline for both co-exposed and independently exposed herds should be no less than eight. The number of reference cattle was determined by gradually increasing the sample size and randomly removing one reference cattle; the change in the median average daily net growth of the herd before and after the random removal should not exceed [a certain threshold]. When that happens, the corresponding sample size is determined as the minimum valid number.

[0102] The normal range for each type of residual is determined by the distribution of residuals from normal samples of the same variety, sex, and growth stage. percentile to the 1st Percentile determination. If the residual exceeds the corresponding normal range and maintains the same direction in two consecutive physiological anchor weighing events, the corresponding residual is judged as abnormal.

[0103] To eliminate the dimensional difference between quality residuals and average daily net growth residuals, the midpoint of the normal range of various residuals is first taken as the zero point, and the original residuals are converted into dimensionless normalized residuals using the half-width of the normal range as the scale. The multi-domain conserved residual fingerprint, expected residual interval, interval boundary distance and overlap length are all determined in the same normalized residual space, and the values ​​of kg and kg / d are not directly compared.

[0104] When an individual's growth residuals are abnormal, but both the common exposure residuals and independent exposure residuals are within the normal range, a candidate source of individual growth abnormality is generated.

[0105] When individual growth residuals and common exposure residuals are abnormal in the same direction within the same statistical period, while independent exposure residuals are within the normal range, a candidate source of common environmental or feed batch abnormality is generated.

[0106] When both commonly exposed and independently exposed cattle herds exhibit measurement consistency residuals in the same direction after passing through the same weighing platform, and the proportion of beef cattle exhibiting this residual in both herds is not less than [percentage missing], At that time, candidate sources of abnormality for the weighing platform are generated. Take the proportion of affected beef cattle from the confirmed abnormal samples of the weighing platform. Percentile.

[0107] When the quality conservation residuals show anomalies in the same direction three times consecutively within the same time window after the same type of feeding, drinking, or excretion event, while the measurement consistency residuals are within the normal range, a candidate source of reversible quality state anomalies is generated.

[0108] S107. Select the maximum separability retest action and confirm the source of the anomaly. The re-testing procedures include re-weighing the target beef cattle at the physiological anchor point on the original weighing platform, cross-weighing the target beef cattle in the backup weighing channel, re-weighing the reference beef cattle on the original weighing platform, increasing the data collection volume of cattle with common exposure and cattle with independent exposure, and performing zero-point verification on the original weighing platform under no-load conditions.

[0109] For each retest action, the integrated early warning controller determines the expected residual direction and expected residual interval corresponding to each candidate abnormality source based on the weighing platform calibration data, dual temporary state calibration data, historical retest records, and confirmed abnormal samples.

[0110] For each candidate source of anomaly, if there are no fewer than 20 confirmed abnormal samples, the expected residual interval is formed by using the 2.5th to 97.5th percentiles of the normalized residuals after retesting; if there are fewer than 20 abnormal samples, the expected residual interval is generated by using a weighing platform calibration test or a single-factor perturbation test with dual temporary state parameters, and marked as a low sample interval, which is then replaced on a rolling basis after new confirmed samples are added.

[0111] The separability of the retest actions is compared in the following order: First, the number of candidate anomaly pairs whose expected residual intervals corresponding to different candidate anomaly sources do not overlap is compared, and the retest actions with more pairs are given priority; if the number of candidate anomaly pairs is the same, the minimum boundary distance between non-overlapping intervals is compared, and the retest actions with larger minimum boundary distances are given priority; if the minimum boundary distances are the same, the total overlap length of all expected residual intervals is compared, and the retest actions with smaller total overlap lengths are given priority; if the total overlap lengths are the same, the number of candidate anomaly pairs with opposite expected residual directions is compared, and the retest actions with more pairs are given priority; if the above results are still the same, the retest actions with shorter execution time, shorter target beef cattle travel distance, and less disturbance to normal feeding are selected in that order.

[0112] After the retest is completed, the integrated early warning controller redetermines the measurement consistency residuals, mass conservation residuals, individual growth residuals, common exposure residuals, and independent exposure residuals to form the retest residual fingerprint.

[0113] If the retested residual fingerprint matches only the expected residual direction and expected residual interval of one candidate anomaly source, the anomaly source is confirmed; if it matches two or more candidate anomaly sources, the matching candidate anomaly sources are retained, and the retesting action with the greatest separability is selected from the retesting actions that have not yet been executed; if it does not match any candidate anomaly source, an unconfirmed anomaly warning is output and the process is entered into manual review.

[0114] The retest matching uses a dimension-by-dimensional determination: each normalized residual actually collected during the retest should fall within the expected residual interval corresponding to the candidate anomaly source, and the residual directions should be consistent; a valid match includes at least the target anomaly residual and two other types of residuals. Residual dimensions not collected during the retest are not included in this matching and are not padded with zeros for missing values.

[0115] S108, Perform selective feedback update When an anomaly is confirmed in the weighing platform, the position-velocity dynamic deviation calibration table is updated only using the results of no-load zero-point calibration, standard mass range calibration, and cross-weighing of the backup weighing channel. The abnormal data is then frozen to update the dual temporary state parameters and the three types of growth baselines.

[0116] When a reversible quality status anomaly is confirmed, only the status retention parameter or the discharge quality deduction ratio corresponding to the source of the anomaly is updated, and the updates of the symmetric weighing platform parameters and the three types of growth baselines are frozen for the abnormal data.

[0117] When an individual growth abnormality is confirmed, an individual growth abnormality warning is issued and the longitudinal baseline of the target beef cattle is frozen; the baselines of the co-exposed herds and the independently exposed herds continue to be updated using valid data from unaffected reference beef cattle.

[0118] When an anomaly is confirmed in the shared environment or feed batch, a corresponding warning is issued and the individual longitudinal baselines of affected beef cattle and the baselines of the co-exposed herds are frozen; the baselines of independently exposed herds continue to be updated using valid data from unaffected reference beef cattle. After corrective measures are taken, when two consecutive physiological anchor periods return to normal, the affected baseline updates resume from the first normal physiological anchor weighing event.

[0119] When the source of the anomaly is not confirmed, it is prohibited to update the weighing platform parameters, dual temporary status parameters, and three types of growth baselines with the abnormal data.

[0120] Example 2:

[0121] like Figure 1 As shown, this embodiment provides a beef cattle growth parameter fusion weighing and early warning system, including an individual identification module, a weighing acquisition module, a growth parameter acquisition module, a quality event acquisition module, a breeding information interface, a fusion early warning control module, a retest execution module, an early warning output module, and a power supply module.

[0122] The individual identification module includes an entry trigger sensor, an electronic ear tag reader, and an identification communication interface. The output of the entry trigger sensor is connected to the start-up terminal of the electronic ear tag reader, and the data output of the electronic ear tag reader is connected to the fusion early warning control module through the identification communication interface.

[0123] The weighing acquisition module includes a weighing platform, eight load detection units, eight signal conditioning circuits, a multi-channel analog-to-digital converter, a unified sampling clock, and a weighing communication interface. The load output terminal of each load detection unit is connected to the input terminal of the corresponding signal conditioning circuit, and the output terminals of the eight signal conditioning circuits are respectively connected to the corresponding input channels of the multi-channel analog-to-digital converter.

[0124] The multi-channel analog-to-digital converter outputs the load value, acquisition time, and load detection unit location identifier to the fusion early warning control module via the weighing communication interface. A unified sampling clock is connected to each of the eight analog-to-digital conversion channels, ensuring that the eight sub-load values ​​are acquired at the same sampling time.

[0125] The growth parameter acquisition module includes a top-view depth camera, a side-view depth camera, an image trigger, an image processing unit, and a posture data interface. The top-view depth camera outputs a depth frame above the weighing platform, and the image processing unit identifies the limb contours, the lowest point of the hoof, and the landing point of the limbs based on the depth frame.

[0126] The side-view depth camera outputs lateral depth frames of beef cattle. The image processing unit identifies body length, body height, chest width, trunk contour volume, trunk pitch angle, trunk lateral tilt angle, and trunk vertical position based on these lateral depth frames. The image trigger is connected to a unified sampling clock to establish a time mapping between the load sampling time and the depth image.

[0127] The quality event acquisition module includes a feeding quality acquisition unit, a drinking water quality acquisition unit, a rumination status acquisition unit, and an excretion event acquisition unit.

[0128] The feeding quality acquisition unit includes a feeding position identifier, a feed trough weighing sensor, and a feeding record interface, which are used to output individual identification, feeding start time, feeding end time, and feeding quality inflow.

[0129] The drinking water quality acquisition unit includes a drinking water level identifier, a flow meter, and a water temperature sensor, which are used to output individual identification, drinking start time, drinking end time, and drinking water quality inflow.

[0130] The rumination status acquisition unit includes a neck motion sensor and a rumination status processor, which are used to output the rumination start time, rumination end time, and rumination duration.

[0131] The excretion event acquisition unit includes a top-down camera of the barn area, a cattle position tracker, a waste collection area quality detection unit, and an event confirmer. The event confirmer outputs the excretion event and excretion quality when changes in excretion posture, cattle position, and waste collection area quality simultaneously meet the corresponding judgment conditions.

[0132] The breeding information interface connects to the breeding management server to read the target beef cattle's barn area identifier, feed batch identifier, environmental control area identifier, growth stage, breed, sex, age in months, and weight upon entry into the pen.

[0133] The integrated early warning and control module includes a clock synchronization unit, an event data packet construction unit, a trusted weighing reconstruction unit, a dual temporary state storage unit, a physiological anchor point calibration unit, a residual fingerprint construction unit, a candidate anomaly generation unit, a retest decision unit, and a selective update unit.

[0134] The clock synchronization unit receives the time stamps from each acquisition module and ensures that the time difference does not exceed [a certain value]. The data is linked at the same time. The event data packet construction unit uses the individual identifier as an index to write the weighing data and growth parameters within the weighing event recording interval into the weighing subset, and writes the quality events and breeding information within the physiological event association interval into the physiological event subset, and links the two subsets by timestamp.

[0135] The reliable weighing reconstruction unit determines the load center according to formula (1), judges whether the load center is located in the effective inner shrinkage area of ​​the supporting polygon, and determines the reliable observation weight according to formula (2).

[0136] The dual temporary storage state unit is based on the quality event sequence number. To update the discrete sequence number, update the rapid temporary storage status of drinking water and the slow temporary storage status of feed according to formula (3); in the first... When a reliable weighing event occurs, the dual temporary states are transferred to the weighing time. and output net growth weight .

[0137] The physiological anchor calibration unit detects the steady-state intervals of feeding and drinking, excretion events, reversible mass change rate, and weighing reliability. When all conditions are met, it generates a physiological anchor identifier and calibrates the dual temporary state parameters in sequence.

[0138] The residual fingerprint construction unit identifies measurement consistency residuals, quality conservation residuals, individual growth residuals, common exposure residuals, and independent exposure residuals. The candidate anomaly generation unit generates candidate sources of individual growth anomalies, common environment or feed batch anomalies, weighing platform anomalies, and reversible quality state anomalies based on residual combination relationships.

[0139] The retesting decision unit compares retesting actions sequentially based on the number of non-overlapping intervals, minimum boundary distance, total overlap length, and expected residual direction difference, and outputs the target retesting instruction. The selective update unit grants update permissions for corresponding parameters based on the confirmed source of the anomaly, while keeping other parameters and the baseline frozen.

[0140] The retest execution module includes a diversion mechanism, a waiting area entrance gate, the original weighing platform entrance gate, a backup weighing channel, a weighing verification device, and a data acquisition node controller. The weighing verification device includes an unloaded zero-point detection unit, a standard mass loading mechanism, and a loading position detector.

[0141] The backup weighing channel uses an independent load detection unit, an independent signal conditioning circuit, and an independent position-velocity dynamic deviation calibration table, without calling the compensation parameters of the original weighing platform.

[0142] The weighing calibration device first continuously weighs the platform After the platform is in an unloaded state, an unloaded zero-point check is performed; if the zero-point offset is within the allowable range, then the standard mass loading mechanism applies pressure to the weighing platform. The standard quality is used for range calibration. lie in to The middle part of the normal weighing range is used to test the response at medium range and is not used for no-load zero-point calibration.

[0143] The data acquisition node controller is used to adjust the data acquisition frequency and the reference beef cattle number for both co-exposed and independently exposed cattle herds. When performing supplemental testing, the data acquisition frequency for the corresponding herd is increased to twice the original frequency, and the reference beef cattle number is increased back to the original number. This lasts for two physiological anchor point cycles, and is repeated twice.

[0144] The early warning output module includes an on-site audio-visual terminal, an early warning interface for the aquaculture management server, and a verification task output interface. It outputs an equipment verification early warning when a weighing platform malfunction is confirmed; an early warning for status parameter calibration when a reversible quality state malfunction is confirmed; an individual verification early warning when an individual's growth is confirmed; and an early warning for the corresponding group when a common environment or feed batch malfunction is confirmed.

[0145] The power supply module includes an AC input terminal, an isolated power supply, a DC regulated power supply, and a backup power supply. The AC input terminal supplies power to the weighing acquisition module, depth camera, fusion early warning control module, and retest execution module via the isolated power supply; the backup power supply maintains the storage of individual identifiers, event data packets, and parameter freeze states when the AC power supply is interrupted.

[0146] The system forms a closed-loop control process of "data acquisition - reliable weighing - reversible mass separation - residual attribution - retest action execution - retest feedback - anomaly confirmation - selective parameter update".

[0147] Example 3:

[0148] like Figure 5 As shown, this embodiment applies the invention to a large-scale cattle farm with an annual stock of 480 head, primarily producing fattening beef cattle. The farm has four feeding units, each accommodating 120 beef cattle.

[0149] The first and second feeding units used feed batches. And connected to the environmental control branch respectively. and environmental control branch The third feeding unit uses batch feed. It is connected to the same environmental control branch as the first feeding unit. The fourth feeding unit uses batch feed. And connect to an independent environmental control branch. .

[0150] This arrangement enables the system to differentiate between feed batch effects and environmental control zone effects by examining cross-exposure relationships between different feeding units.

[0151] The weighing platform is located in a one-way passage from the watering area back to the fattening shed, and the beef cattle pass through the weighing platform naturally once or twice a day. The weighing results are written into the reliable observation weight record only when the load-support geometry and motion consistency conditions specified in Example 1 are met.

[0152] In a 14-day statistical period, the individual identifier is The credible observed weight of the target beef cattle in the previous physiological anchor weighing event was The corresponding drinking water rapid temporary storage quality is The slow-storage quality of feed is According to formula (3), the net growth weight corresponding to the previous physiological anchor weighing event is: .

[0153] In the current physiological anchor weighing event, the reliable observed weight of the target beef cattle is: The corresponding drinking water rapid temporary storage quality is The slow-storage quality of feed is Net growth weight is .

[0154] Net growth weight increase between two adjacent physiological anchor weighing events The corresponding average daily net growth is .

[0155] The individual longitudinal baseline of the target beef cattle at its current growth stage is: The baseline net increase for the fourteenth day was Therefore, the current net growth weight deviation relative to the individual's longitudinal baseline is negative. .

[0156] The median average daily net increase in the common exposed herds of the target beef cattle during the statistical period was: The median historical baseline of this cattle herd is The residuals are exposed together. It shows a negative anomaly; the median average daily net growth of independently exposed cattle herds during this statistical period was The median historical baseline of this cattle herd is Independent exposure residuals are It is within the permissible range.

[0157] The measurement consistency residuals for the target beef cattle were: The residual value is within the normal range; the mass conservation residual is... This is also within the normal range. Based on this, the fusion early warning controller generates candidate sources of common environmental or feed batch anomalies.

[0158] To differentiate between environmental control branch anomalies and feed batch anomalies, the retesting decision unit compared the expected residual intervals corresponding to each retesting action. Since the first and second feeding units used the same feed batch... However, they are connected to the environmental control branch respectively. and environmental control branch The third rearing unit is connected to the first rearing unit via the same environmental control branch. However, the feed batches used Therefore, adding reference beef cattle samples from three feeding units can distinguish between two candidate sources of abnormality.

[0159] The integrated early warning controller increases the number of reference beef cattle participating in the comparison in the first, second, and third feeding units to twenty each, and continuously collects data for two physiological anchor point cycles.

[0160] The results showed that sixteen reference beef cattle in the first feeding unit exhibited growth residuals in the same direction, fifteen reference beef cattle in the second feeding unit exhibited growth residuals in the same direction, and only two reference beef cattle in the third feeding unit exhibited growth residuals.

[0161] The first and second feeding units are connected to different environmental control branches, but use the same batch of feed. The third feeding unit is connected to the same environmental control branch as the first feeding unit, but uses different feed batches. Therefore, the retested residual fingerprint matches the expected residual changes corresponding to the feed batch anomaly, confirming that the anomaly originated from the feed batch. .

[0162] The early warning output module outputs an alert for feed batch anomalies and records the individual identifier of the affected cattle, their feeding unit, feed batch identifier, the start time of the anomaly, and the retest results. The integrated early warning controller freezes the individual longitudinal baselines of the affected cattle and the baselines of the commonly exposed herds, without modifying the position-velocity dynamic deviation calibration table and dual temporary state parameters of the weighing platform; the baselines of independently exposed herds continue to be updated using valid data from unaffected reference cattle.

[0163] After changing the feed batch, the system continued to weigh the affected beef cattle at physiological anchor points. The common exposure residuals of the first and second feeding units for two consecutive physiological anchor point cycles returned to the normal range, and the average daily net gain of the target beef cattle returned to normal. .

[0164] After manual review revealed no individual diseases, zero-point offset on the weighing platform, or abnormal water metering, and the affected herd returned to normal for two consecutive physiological anchor point cycles, the fusion early warning controller resumed updating the longitudinal baseline of the affected individuals and the baseline of the co-exposed herd from the first normal physiological anchor point weighing event after the feed batch change. Data from abnormal periods were retained in the abnormality file but were not included in subsequent calculations of normal growth baselines.

[0165] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A beef cattle growth parameter fusion weighing and early warning system, characterized in that, It includes an individual identification device, a weighing platform, a load detection unit arranged in sections on the weighing platform, a limb landing point acquisition device, a trunk posture and body size acquisition device, a quality event acquisition device, an aquaculture information interface, a retest execution mechanism, an early warning output device, and a fusion early warning controller connected to the aforementioned devices. The fusion early warning controller will generate an individual multi-source event data package from the target beef cattle’s individual identifier, load sequence within the weighing event recording interval, limb landing point, trunk posture and body size parameters, as well as feed quality inflow, water quality inflow, excretion events and breeding information within the physiological event interval associated with the current weighing event, according to a unified time reference. The load center is determined based on the load sequence, and a support polygon is constructed based on the landing points of the limbs. When the load center is located in the effective inner area of ​​the support polygon, the total load sequence of the velocity compensation is determined by the load center position and the longitudinal displacement of the load center and the torso center of mass. The reliable observed weight is obtained based on the consistency of the compensation load, the load center trajectory and the torso vertical acceleration. Establish a dual temporary state, use physiological anchor weighing events to correct the dual temporary state and separate the reversible mass from the reliable observed weight to obtain the net growth weight; Based on the credible observation weight, structural constraint weight, reversible mass change, individual longitudinal baseline, and the historical baselines of both the jointly exposed and independently exposed cattle herds, a normalized multi-domain conserved residual fingerprint is formed. Candidate anomaly sources are generated, and the retesting action with the highest separability of the expected residual interval is selected. The retesting execution mechanism is controlled to execute the retesting and obtain the retesting residual fingerprint. Based on the retest residual fingerprint, the source of the anomaly is identified and an early warning is issued. The corresponding parameter update channel is opened or the affected baseline is frozen according to the anomaly type. If the source of the anomaly is not identified, the anomaly data update is frozen.

2. The beef cattle growth parameter fusion weighing and early warning system according to claim 1, characterized in that, The individual identification device is located at the entrance side of the weighing platform; the load detection unit outputs the load value associated with the acquisition time and the longitudinal and transverse positions according to a unified sampling clock; The limb landing point acquisition device includes a top-view depth camera located above the weighing platform, and the torso posture and body size acquisition device includes a side-view depth camera located to the side of the weighing platform. The quality event acquisition device includes a weighing feeding trough with an individual identifier, a metering waterer with an individual identifier, an excrement quality detection unit, and a cattle body position tracking unit; The breeding information interface outputs the target beef cattle's barn area identifier, feed batch identifier, environmental control area identifier, and growth stage; the fusion early warning controller writes the weighing and posture data within the weighing event recording interval into the weighing subset, and writes the quality events within the physiological event interval, which ends with the current weighing event and covers the most recent concentrated feeding, concentrated drinking, and subsequent excretion events, into the physiological event subset, and associates the two subsets according to the individual identifier and timestamp.

3. The beef cattle growth parameter fusion weighing and early warning system according to claim 1, characterized in that, The fusion early warning controller uses the load value of each load detection unit as the load weight of the corresponding bearing position to determine the longitudinal position and lateral position of the load center, and forms the load center trajectory according to the acquisition time; it selects at least three landing points that are currently in an effective support state from the left forelimb, right forelimb, right hindlimb and left hindlimb. A supporting polygon is formed by connecting the circumference of the beef cattle's torso, and the boundary of the supporting polygon is offset inward by a preset safe distance to form an effective inward shrinkage area; The preset safety distance is determined based on the sampling error of the limb landing point, the hoof width, and the width of the load-bearing influence of the edge of the weighing platform; sampling times where the load center is outside the effective inner contraction area are not included in the calculation time set of the reliable observation weight.

4. The beef cattle growth parameter fusion weighing and early warning system according to claim 2, characterized in that, The position-velocity dynamic deviation response relationship is calibrated using a standard mass load at different load positions and moving speeds. This calibration is only used to characterize the position-velocity system deviation of the weighing platform. The passing speed is determined by the longitudinal displacement of the load center and the torso center of mass within a preset time window; the torso vertical acceleration is obtained by the second-order difference of the filtered torso vertical position; the fusion early warning controller compensates for the total load according to the corresponding deviation, and excludes hoof impact sampling based on the compensation load fluctuation, the continuity of the load center trajectory and the vertical acceleration-load mapping relationship. When all consistency conditions are met, the median of the weights corresponding to the compensated loads within the effective bearing range is determined as the reliable observed weight.

5. The beef cattle growth parameter fusion weighing and early warning system according to claim 3, characterized in that, The rapid temporary storage state of drinking water records the water mass that has not yet been reduced through distribution or excretion after drinking, while the slow temporary storage state of feed records the feed mass that has not yet been reduced through digestion, rumination or excretion after feeding. When a drinking event is detected, the water inflow is increased to a rapid temporary storage state; when a feeding event is detected, the feed inflow is increased to a slow temporary storage state. When an excretion event is detected, the amount of deduction in excretion quality between the two temporary states is determined based on the time interval between the excretion event and the most recent drinking and feeding events, the current rumination status, and the growth stage. The two temporary storage states are transferred according to the event interval and the state retention parameters corresponding to the growth stage. The parameters of the rapid water storage state are less than those of the slow feed storage state. The transfer results are taken as non-negative values. Event intervals exceeding 4 hours are processed in segments. Background mass loss that is not observed separately is included in the state retention parameters.

6. The beef cattle growth parameter fusion weighing and early warning system according to claim 4, characterized in that, The fusion early warning controller determines the current weighing event as a physiological anchor weighing event when the target beef cattle have reached a stable feeding interval since the most recent concentrated feeding, a stable drinking interval since the most recent concentrated drinking, an excretion event has been detected after the most recent concentrated feeding, the rate of change of the sum of the two temporary states is within a stable range, and a reliable observed weight is obtained for the current weighing event. Based on the reliable observed weight changes between adjacent physiological anchor weighing events, and the mass balance relationship formed by the inflow of feed quality, the inflow of drinking water quality, and the outflow of excretion quality within the corresponding time interval, the state retention parameters of the two temporary states and the excretion quality deduction ratio are corrected according to the mass balance relationship. The net growth weight is obtained by subtracting the corrected rapid water storage mass and slow feed storage mass from the current reliable observed weight.

7. The beef cattle growth parameter fusion weighing and early warning system according to claim 5, characterized in that, The body size-weight mapping relationship is a lookup table established by dividing the body length, body height, chest width and trunk contour volume into partitions. Each partition records the median static weight and the allowable range. The fusion early warning controller interpolates adjacent zones and corrects them according to the median of the attitude residuals corresponding to the torso pitch angle and lateral tilt angle to obtain the structural constraint weight; The multi-domain conserved residual fingerprint includes measurement consistency residual, mass conservation residual, individual growth residual, common exposure residual, and independent exposure residual; wherein, the common exposure residual and the independent exposure residual are the median growth residuals of each reference beef cattle in the corresponding herd relative to the herd's own historical baseline, respectively. The deviation of each original residual from the midpoint of the normal range is divided by the corresponding half-width to form a dimensionless normalized residual. The fusion early warning controller records the fingerprint according to the direction, amplitude, duration, and time relationship with the quality event of each normalized residual.

8. The beef cattle growth parameter fusion weighing and early warning system according to claim 6, characterized in that, The co-exposed herd consists of reference beef cattle that share barn areas, feed batches, and environmental control areas with the target beef cattle. The independent exposed herd consists of reference cattle that are matched with the target beef cattle in terms of breed, sex, age, entry weight and growth stage, and do not share the barn area or feed batch. When an individual's growth residual exceeds its allowable range, while the co-exposure residual and independent exposure residual are within their respective allowable ranges, a candidate source of individual growth abnormality is generated. When individual growth residuals and common exposure residuals exceed their respective allowable ranges in the same direction within the same statistical period, while independent exposure residuals remain within the allowable range, a common environmental or feed batch anomaly candidate source is generated; when two herds passing through the same weighing platform both exhibit measurement consistency residuals in the same direction that exceed the allowable range, a weighing platform anomaly candidate source is generated; when mass conservation residuals repeatedly exceed limits in the same direction within a preset time window following the same type of feeding event, drinking event, or excretion event, while measurement consistency residuals remain within the allowable range, a reversible mass state anomaly candidate source is generated.

9. The beef cattle growth parameter fusion weighing and early warning system according to claim 7, characterized in that, The retesting execution mechanism includes a diversion mechanism connected to the outlet of the original weighing platform, a backup weighing channel independent of the original weighing platform, a weighing verification device, and a data acquisition node controller; the weighing verification device includes an unloaded zero-point detection unit and a standard mass loading mechanism. The re-testing actions include re-weighing the target beef cattle at the physiological anchor point on the original weighing platform, cross-weighing in the backup weighing channel, re-weighing the reference beef cattle on the original weighing platform, increasing the collection frequency and the number of reference beef cattle for the two exposed cattle groups, and sequentially performing no-load zero-point verification and standard mass range verification on the original weighing platform. The fusion early warning controller determines the expected residual interval under each retest action in a unified dimensionless normalized residual space; when there are fewer than 20 confirmed abnormal samples, the expected residual interval is generated based on the equipment calibration data and the disturbance results of the state parameters; the separability is determined based on the number of non-overlapping intervals, the minimum boundary distance, the amount of overlap and the residual direction, and the largest one is selected.

10. The beef cattle growth parameter fusion weighing and early warning system according to claim 9, characterized in that, The fusion early warning controller controls the diversion mechanism, backup weighing channel, weighing verification device or data acquisition node controller to execute the selected retest action, and forms a retest residual fingerprint based on the five types of normalized residuals after the retest. When the collected dimensions of the retested residual fingerprint fall into the corresponding expected residual interval and the directions are consistent, a match is determined. The matching dimensions include at least the target residual and two other types of residuals. When only one candidate anomaly source is matched, the source is confirmed. When there are more than two matching results, the action with the highest separability among the unexecuted retest actions is executed. When all are unmatched, an unconfirmed anomaly warning is output. When a weighing platform malfunction is confirmed, only the position-velocity dynamic deviation response relationship is updated; when a reversible mass state malfunction is confirmed, only the corresponding temporary state parameters are updated; in both cases, the updates of other parameters and the three types of baselines by abnormal data are frozen; when an individual growth malfunction is confirmed, the longitudinal baseline of the target beef cattle is frozen; when a common environment or feed batch malfunction is confirmed, the longitudinal baseline of the affected individuals and the baseline of the commonly exposed herd are frozen; the baseline of the unaffected group continues to be updated, and the update of the affected baseline is resumed after two consecutive normal physiological anchor point cycles.

Images