A method for controlling the average weight of yellow-feathered capons in large-scale poultry farming

By combining smart chicken scales with Bluetooth chicken scales for data acquisition and processing, the accuracy and stability issues of calculating the average weight of yellow-feathered capons have been resolved. This enables sample compensation and data optimization in high-weight ranges under stress-free conditions, adapting to the needs of different breeding scenarios and growth stages, and improving data reliability and management efficiency.

CN123115586APending Publication Date: 2026-11-17WENS FOODSTUFF GROUP CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611577849.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-13
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In large-scale farming of yellow-feathered capons, existing technologies struggle to improve the accuracy of average weight calculation and data stability in long-term age groups of large, low-activity flocks without stress-induced automatic weighing. In particular, the average weight calculation deviation is significant due to insufficient sample representativeness and skewed sample distribution in the high-weight range.

Method used

The system continuously collects data using a smart chicken scale and combines it with manual sampling using a Bluetooth chicken scale under dark conditions at night. Through weighted correction and joint prediction, combined with standard growth curves and historical average weight data, the system screens samples and assesses confidence levels, dynamically adjusts weight coefficients and fluctuation ratios, compensates for missing samples in the high-weight range, and forms a data optimization closed loop.

Benefits of technology

While maintaining the advantages of stress-free automatic weighing, it significantly improves the accuracy and data stability of average weight calculation, reduces calculation deviation, adapts to the needs of different breeding scenarios and growth stages, and ensures the reliability of refined management and decision-making.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123115586A_ABST
    Figure CN123115586A_ABST
Patent Text Reader

Abstract

This application relates to the technical field of automated poultry farming, and discloses a method for processing the average weight of large-scale poultry flocks of yellow-feathered capons. The method includes: acquiring daily valid samples continuously collected by a smart chicken scale, and manually sampled samples obtained by weighing each chicken individually using a Bluetooth scale under dark conditions at night, with a number exceeding a preset threshold; pre-storing a standard growth curve for yellow-feathered capons and accumulating historical average weight data corresponding to each age group; dividing the two types of samples into multiple equal intervals according to weight, comparing the differences in proportion to determine weight coefficients, and weighting and correcting to obtain the algorithmic average weight; when the sample peak is not obvious, jointly estimating based on the standard growth curve and historical average weight data, recalculating the average weight after screening samples; and conducting a confidence assessment based on the rationality of daily average weight gain and the variation range of the coefficient of variation. This application can improve the accuracy of average weight calculation and the data stability of long-term poultry flocks while maintaining the advantages of stress-free automatic weighing.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the technical field of automated breeding, and in particular to a method for weight processing of large-scale poultry flocks of yellow-feathered capons. Background Technology

[0002] Yellow-feathered capons have a long rearing period and a large slaughter weight. Their average weight and uniformity are core indicators for evaluating breeding efficiency, determining the timing of slaughter, and influencing market pricing power. Currently, large-scale chicken houses commonly use in-pen smart chicken scales to achieve 24 / 7 stress-free automatic weighing. Relying on the Internet of Things, they continuously collect weight data and calculate the average weight of the flock, thus replacing the traditional method of manually weighing chickens, reducing stress on the flock, and improving monitoring efficiency.

[0003] For large-sized, low-activity, long-lived breeds like yellow-feathered capons, there are significant differences in weight among individuals, and chickens in the high-weight range are less willing to be weighed. The daily effective sample size collected by a single smart chicken scale is limited, and the sample weight distribution is systematically skewed from the actual weight distribution of the flock. In particular, the peak value of the sample distribution is not obvious in the long-lived, high-weight stage. The results obtained by using conventional average weight calculation methods have large deviations and insufficient data stability, making it difficult to support refined feeding management and sales decisions.

[0004] As can be seen from the above, how to improve the accuracy of average weight calculation and the data stability of large-sized, low-activity chicken flocks while maintaining the advantages of stress-free automatic weighing is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] In order to improve the accuracy of average weight calculation and the data stability of large-sized, low-activity chicken flocks while maintaining the advantages of stress-free automatic weighing, this application provides a method for processing the average weight of large-scale chicken flocks suitable for yellow-feathered capons.

[0006] Firstly, this application provides a method for weight management of large-scale poultry flocks of yellow-feathered capons, employing the following technical solution: A method for controlling the average weight of yellow-feathered capons in large-scale poultry farming includes: The system acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken individually using the Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than the preset threshold. The system also pre-stores the standard growth curve of yellow-feathered capons and accumulates and stores the historical average weight data corresponding to each feeding age. The daily valid samples and manually sampled samples are divided into multiple equal intervals according to weight. The sample proportion in each equal interval is calculated and the difference between the two proportions is compared. The weight coefficient of each equal interval is determined based on the difference in proportion. The daily valid samples are then weighted and corrected using the weight coefficient to obtain the algorithm average weight. When the daily effective sample meets the condition that the peak is not obvious, a joint prediction is made based on the pre-stored standard growth curve and the accumulated historical average weight data. The average weight is then recalculated after screening the daily effective samples according to the joint prediction results. The confidence level of the calculated average weight is evaluated based on the reasonableness of the daily average weight gain obtained from the algorithmic average weight over several consecutive days, and the variation range of the coefficient of variation obtained from the statistics of effective samples.

[0007] Optionally, the steps for obtaining daily valid samples continuously collected by the smart chicken scale and manually sampled samples include: A smart chicken scale with an extended weighing pan is used to collect data continuously throughout the day. The collected data is processed to remove jitter, identify multiple chickens, and remove outliers to obtain a daily valid sample. In the dark at night, the flock of chickens is delineated centered on the location where the smart chicken scale is placed. Each chicken is weighed individually using the Bluetooth chicken scale to obtain a manually sampled sample. The manually sampled sample covers three weight ranges: small weight, medium weight, and large weight.

[0008] Optionally, the step of dividing the daily valid samples and manually sampled samples into multiple equal intervals according to weight and then weighting and adjusting them to obtain the algorithm's average weight includes: The daily valid samples and manually sampled samples were divided into multiple equal intervals according to weight from smallest to largest, and the sample proportion in each interval was calculated. By comparing the differences in the proportion of the two types of samples in the corresponding intervals, the weight coefficient is increased for the high-weight intervals in the daily valid samples that have a lower proportion than the manually sampled samples, and the weight coefficient is decreased for the intervals that have a higher proportion than the manually sampled samples. After normalizing the weight coefficients of each interval, the average weight of the algorithm is obtained by weighting the effective samples of each day.

[0009] Optionally, the steps of performing joint prediction and recalculating the average weight when the daily effective sample meets the condition of no significant peak include: When the peak of the weight distribution of the daily effective samples is not obvious, a preliminary estimate is obtained based on the pre-stored standard growth curve of yellow-feathered capons and the current age of the chickens. Time series prediction values ​​are obtained by performing time series smoothing based on accumulated historical average weight data. Based on the preliminary and time-series estimates, the upper and lower limits of sample selection are determined. After filtering out abnormal samples from the daily valid samples, the average weight is recalculated.

[0010] Optionally, the process of manually sampling chickens by weighing them one by one using a Bluetooth chicken scale includes: Representative pens were manually sampled in both environmentally controlled and non-environmentally controlled chicken houses, and weighing data were recorded under different conditions. If the number of samples in a single manual sampling exceeds a preset threshold, manual sampling will be repeated at fixed intervals to update the baseline proportion of each weight range under the corresponding environment.

[0011] Optionally, the steps of determining the upper and lower limits of sample selection based on the preliminary estimate and the time-series estimate include: The baseline average weight value is calculated by assigning corresponding weights to the preliminary estimate and the time series estimate, respectively. The floating ratio is determined based on the current growth stage of the animals at their current age, and the upper and lower limits of sample selection are calculated based on the baseline average weight. The longer the animal is kept, the smaller the corresponding fluctuation ratio.

[0012] Optionally, the step of evaluating the confidence level of the calculated algorithm average weight includes: The daily average weight gain calculated from the algorithm's average weight over several consecutive days and the variation range of the coefficient of variation obtained from the statistical analysis of the effective samples on the day are verified separately. Based on the results of the two verifications, the algorithm's average weight is divided into different confidence levels. When the confidence level is lower than the preset level, the corresponding abnormality flag is output and the corresponding level of review reminder is triggered.

[0013] Optionally, it includes: When the confidence level is lower than the preset level, supplementary manual sampling is carried out for the corresponding enclosure under dark conditions at night. The sampling focuses on covering the higher weight range. The samples obtained from the supplementary sampling are used to update the weight range ratio benchmark under the corresponding environment, and the weight coefficient of each equal interval is re-determined. The average weight of the algorithm for the day is then corrected a second time before being output.

[0014] Optionally, it includes: When determining the floating ratio, first match the base ratio to the environmental type corresponding to the current pen, and then adjust the base ratio based on the growth stage of the current age of the animals and the coefficient of variation of the effective samples on the day. The higher the age of the animals, the smaller the floating ratio, and the larger the coefficient of variation, the larger the floating ratio.

[0015] Secondly, this application provides a system for weight processing of large-scale poultry flocks of yellow-feathered capons, employing the following technical solution: A system for weight management in large-scale poultry farming of yellow-feathered capons, comprising: The sample acquisition module acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken in the flock via Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than the preset threshold. The standard growth curve of yellow-feathered capons is pre-stored and the historical average weight data corresponding to each feeding age is accumulated and stored. The interval weighted correction module divides the daily valid samples and manually sampled samples into multiple equal intervals according to weight, calculates the sample proportion in each equal interval and compares the difference between the two proportions, determines the weight coefficient of each equal interval based on the proportion difference, and uses the weight coefficient to perform weighted correction on the daily valid samples to obtain the algorithm average weight. The sample joint screening module performs joint prediction based on the pre-stored standard growth curve and the accumulated historical average weight data when the daily effective samples meet the condition of no obvious peak. After screening the daily effective samples according to the joint prediction results, the average weight is recalculated. The confidence assessment module assesses the confidence of the calculated average weight based on the reasonableness of the daily average weight gain obtained from the algorithm's average weight over multiple consecutive days, and the variation range of the coefficient of variation obtained from the statistics of valid samples.

[0016] In summary, this application includes at least one of the following beneficial technical effects: While retaining the advantages of intelligent chicken scales for automatic weighing without stress around the clock, the true weight distribution benchmark of the flock is obtained through manual sampling under dark conditions at night. Combined with the comparison of weight interval proportions and dynamic weight adjustments, the missing samples in the high weight interval are specifically compensated, effectively correcting the deviation in average weight calculation caused by sample distribution skewness. To address the issue of indistinct peak values ​​in the sample distribution during the long-term, high-weight stage, joint estimation is carried out based on pre-stored standard growth curves and accumulated historical data. This is used to define reasonable sample screening boundaries to filter out abnormal data, ensuring the stability of average weight calculation during the long-term stage and improving the overall accuracy of average weight calculation for large, low-activity flocks.

[0017] Based on this, the design further adapts to the application needs of different breeding scenarios and growth stages by sampling and calibration in environmentally controlled and non-environmentally controlled environments, dynamically adjusting the screening floating ratio according to the age of the animals, and dividing confidence levels with dual-dimensional indicators. When the data confidence is insufficient, secondary correction can be completed by supplementary sampling covering the high weight range, forming a complete data optimization closed loop. Under the premise of maintaining low-stress breeding throughout the process, the reliability of average weight data and the reference value of breeding management are continuously guaranteed. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an exemplary embodiment of a method for processing the average weight of a large-scale poultry flock of yellow-feathered capons.

[0019] Figure 2 This is a structural block diagram illustrating a weight equalization system for large-scale breeding flocks of yellow-feathered capons, according to an exemplary embodiment. Detailed Implementation

[0020] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0021] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] This application discloses a method for weight management in large-scale yellow-feathered capons, primarily applicable to farms with over 10,000 capons, covering various farming scenarios including environmentally controlled enclosed houses and semi-open, non-environmentally controlled houses. Yellow-feathered capons are typically raised for over 120 days, with slaughter weights mostly between 2.8 kg and 3.5 kg. Significant weight differences exist between individuals, and as the age of the capons increases, the activity level of larger birds decreases, leading to a substantial drop in their willingness to use smart scales. In current large-scale farming scenarios, fixed smart chicken scales are commonly used in pens for automatic weighing and monitoring. However, for large-sized, low-activity, long-lived breeds like yellow-feathered capons, conventional smart chicken scales generally suffer from insufficient sample representativeness: chickens in the high-weight range are weighed less frequently, resulting in a systematic skew between the daily sample weight distribution and the actual weight distribution of the flock, leading to an overall lower average weight calculation result. Furthermore, during the long-lived stage before slaughter, the weight gain of the flock slows down, the sample distribution becomes more dispersed, and the distribution peak becomes blurred. Conventional average weight calculation methods exhibit large data fluctuations and poor stability, with the calculated results often deviating from the actual average selling weight by more than 100g. This makes it difficult to support refined feeding adjustments, accurate slaughter timing determination, and market pricing decisions, and also fails to meet the digital management needs of large-scale farming.

[0023] In this application embodiment, a method for weight processing of large-scale poultry flocks of yellow-feathered capons is described, referring to... Figure 1 It includes the following steps: S100 acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken individually using the Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than a preset threshold. The standard growth curve of yellow-feathered capons is pre-stored and the historical average weight data corresponding to each feeding age is accumulated and stored.

[0024] In this embodiment of the invention, the specific execution process of S100 is as follows: S101, Smart Chicken Scale for Data Acquisition and Valid Sample Acquisition. A smart chicken scale with an extended weighing pan is deployed near the drinking line or feeding passage in the capon rearing pens. The pan length is more than 30% longer than conventional smart chicken scales, the width is adapted to the body size of yellow-feathered capons, and the pan height is reduced by 15% to 20% to minimize obstacles for the chickens as they lift their legs, accommodating the natural standing and walking posture of large capons. This physical structure improves the chickens' willingness to be weighed and the daily weighing rate. The smart chicken scale is connected to the livestock IoT system for continuous 24 / 7 data acquisition. The raw weight data acquisition frequency is set to twice per second to ensure that the complete weighing process of the chickens is captured. The collected data is uploaded in real time to an edge computing gateway or livestock cloud platform via LoRa or Ethernet.

[0025] The collected raw weight data underwent sequential processing including vibration removal, multiple chicken identification, and outlier removal to obtain daily valid samples. Vibration removal was achieved by identifying the frequency and amplitude of weight data fluctuations. When weight values ​​exhibited high-frequency, small-amplitude fluctuations within a short period, they were identified as invalid data generated by chicken movement and filtered out, retaining only values ​​from periods of stable weight as valid weighing data. Multiple chicken identification was based on comparing weight values ​​with the weighing pan's load-bearing threshold. When a weight value was detected exceeding 1.8 times the reasonable weight range for a single chicken, and the weight fluctuation characteristics matched the superposition of multiple chickens weighing at different times, it was identified as a scenario of multiple chickens weighing simultaneously, and this data segment was removed to avoid data distortion caused by the superposition of multiple chicken weights. Outlier removal employed the 3σ criterion. Based on all preliminary valid weighing data for the day, the mean and standard deviation were calculated. Data points exceeding the mean ± 3 times the standard deviation were identified as outliers and removed. Simultaneously, a secondary verification was performed using the reasonable weight range for the corresponding age of yellow-feathered capons; extreme data that clearly deviated from the growth pattern were also removed. After triple processing, a stable and reliable set of valid daily samples is finally obtained, which is used for subsequent average weight calculation.

[0026] S102, Large-scale manual sampling under dark conditions. Representative pens were selected in both environmentally controlled and non-environmentally controlled chicken houses for manual sampling. Environmentally controlled chicken houses have stable temperature and humidity and uniform feeding conditions, resulting in relatively low weight dispersion of chickens. Non-environmentally controlled chicken houses are greatly affected by the external environment, resulting in greater weight differences among chickens. Therefore, sampling benchmarks were established for each environment to improve adaptability to different scenarios.

[0027] Manual sampling was conducted 1 to 2 hours after lights out at night, when the chickens were quiet and resting with minimal stress. Sampling was carried out by gradually enclosing the surrounding flock using a fence, centered on the location of the smart chicken scale, to avoid large-scale herding and frightening. Each chicken was weighed individually using a portable Bluetooth scale, and the weight data was automatically synchronized to the handheld terminal and the breeding platform, eliminating the need for manual recording and improving sampling efficiency and data accuracy. The manual sampling covered three weight ranges: low, medium, and high. These ranges were defined according to the standard weight range for their corresponding age: below 90%, 90% to 110%, and above 110% of standard weight, ensuring that the sampling covered the entire weight range of the flock and accurately reflected their weight distribution.

[0028] The sample size for each manual sampling is greater than a preset threshold, which is set based on the number of birds in the pen, typically not less than 5% of the total number of birds, and the absolute number is not less than 150 birds, to ensure that the sampling results are statistically representative. Manual sampling is repeated at fixed intervals: once every 7 days in the early growth stage, once every 3 days in the middle and late growth stages, and once every 2 days one week before slaughter. The baseline proportions of each weight range under the corresponding environment are updated synchronously to ensure that the baseline data matches the growth stage of the flock and avoid errors caused by the lag in the distribution baseline due to the growth of the flock.

[0029] S103, Basic Data Pre-storage and Cumulative Storage. The standard growth curve for the corresponding breed of yellow-feathered capon is pre-stored. This curve is derived by fitting the official feeding standards for this breed with historical growth data from large-scale farms. It records the theoretical weight, average daily weight gain reference value, and weight distribution reference coefficient for different feeding ages, serving as a benchmark for growth stage judgment and data verification. The standard growth curve refines data nodes by age week. In practical applications, linear interpolation is used to obtain reference values ​​for any feeding age.

[0030] Historical average weight data for each pen and age of rearing is stored cumulatively. Each data record includes core fields such as the corresponding date, age of rearing, pen number, environment type, effective sample size for the day, average weight according to the algorithm, sample coefficient of variation, and average daily weight gain. Historical data is updated daily and regularly cleaned to remove distorted data from special scenarios such as equipment failure and abnormal weather, ensuring the reliability of the historical dataset. The accumulated historical average weight data provides comprehensive data support for subsequent time series prediction, coefficient of variation comparison, and confidence assessment.

[0031] S200 divides the daily valid samples and manually sampled samples into multiple equal intervals according to weight, calculates the sample proportion in each equal interval and compares the difference in proportion between the two, determines the weight coefficient of each equal interval based on the difference in proportion, and uses the weight coefficient to weight and correct the daily valid samples to obtain the algorithm average weight.

[0032] In this embodiment of the invention, the specific execution process of S200 is as follows: S201, Weight Distribution Divided into Ten Equal Intervals. The daily valid samples and manually sampled samples were sorted by weight from smallest to largest. Using the minimum and maximum weight values ​​of the day as upper and lower boundaries, the sample was divided into 10 equal intervals, each with an equal weight width. The number of samples and their corresponding proportion within each interval were calculated. Weight distribution curves for both types of samples were generated with the interval number on the horizontal axis and the sample proportion on the vertical axis.

[0033] The interval division adopts a dynamic equal division mechanism. When the effective sample size of a day is lower than the preset sample size threshold, the number of intervals is appropriately reduced to 7 to 8 to avoid statistical errors caused by insufficient sample size within each interval. When the sample size is sufficient, the division accuracy of 10 intervals is maintained to ensure the resolution of distribution features. By dynamically adjusting the number of intervals, both statistical stability and distribution granularity under different sample size scenarios are taken into account.

[0034] S202, Distribution Difference Comparison and Skewness Calculation. The distribution curve of the daily valid samples from the smart chicken scale is compared with the distribution curve of the manually sampled samples during the same period, one interval at a time. The difference in sample proportion for each interval is calculated, i.e., the manual sampling proportion minus the daily valid sample proportion. The distribution of these proportion differences identifies the skewness characteristics of the smart chicken scale samples relative to the true distribution: when the difference in proportion for the higher weight interval is positive and large, it indicates that the proportion of high-weight chickens in the smart chicken scale sample is significantly insufficient, and the overall sample distribution is skewed towards lower weight; when the difference in proportion for the lower weight interval is positive, it indicates that the sample size in the lower weight interval is insufficient.

[0035] The study focuses on quantitatively analyzing the differences in the proportion of high-weight individuals, calculating the cumulative difference in the proportion of high-weight individuals as an indicator of sample distribution skewness. A larger value indicates more severe missing high-weight samples and a higher risk of bias in the average weight calculation. By quantifying skewness characteristics, a clear basis is provided for subsequent weight adjustments, enabling targeted correction.

[0036] S203, Interval Weight Adjustment and Average Weight Correction Calculation. The corresponding weight coefficient is calculated based on the percentage difference of each interval. The adjustment of the weight coefficient is positively correlated with the percentage difference: for intervals with a higher weight percentage in the daily valid samples than in the manually sampled samples, the larger the percentage difference, the greater the increase in the weight coefficient; for intervals with a higher weight percentage than in the manually sampled samples, the percentage difference is negative, and the corresponding weight coefficient is decreased. The initial baseline value of the weight coefficient is 1. The ratio of the adjustment magnitude to the percentage difference is set according to the farming scenario and breed characteristics to ensure that the corrected distribution matches the true distribution.

[0037] The weight coefficients for all intervals are normalized, and the sum of all weight coefficients is adjusted to match the number of intervals. This ensures that the total number of samples after weighted calculation is equivalent and avoids deviations in the average weight calculation caused by overall weight shift. After normalization, each data point in the daily valid samples is weighted according to the weight coefficient corresponding to its interval, and then the average of all weighted data is calculated to obtain the corrected average weight of the algorithm.

[0038] By using interval weighting correction, the distribution skew caused by insufficient samples in the high-weight interval can be specifically compensated. Compared with the calculation method of directly using the original sample mean, it is closer to the actual weight distribution of the flock and solves the problem of low average weight caused by low weighing rate of large-sized chickens from the distribution level.

[0039] S300, when the daily effective sample meets the condition of no obvious peak, performs joint prediction based on the pre-stored standard growth curve and the accumulated historical average weight data, and recalculates the average weight after screening the daily effective samples according to the joint prediction results.

[0040] In this embodiment of the invention, the specific execution process of S300 is as follows: S301, Determination of Indistinct Peak Values ​​and Triggering of Joint Prediction. The kurtosis coefficient of the body weight distribution of valid samples for the day is calculated. The kurtosis coefficient reflects the sharpness of the peak value of the distribution curve; the lower the value, the smoother the distribution and the less distinct the peak value. When the kurtosis coefficient is lower than a preset threshold, or the proportion of samples in the largest frequency range of the distribution is lower than a preset proportion, the sample distribution is determined to have indistinct peak values, triggering the joint prediction and sample screening process for the long-term, high-weight stage.

[0041] To avoid false triggering due to daily data fluctuations, a two-day judgment mechanism is adopted: joint prediction screening is only formally initiated when the sample distribution meets the condition of no significant peak for two consecutive days; if only a single day shows no significant peak, it is not triggered and only marked for observation. This judgment mechanism is adapted to the characteristic of gradually increasing sample dispersion during the long-term growth stage of yellow-feathered capons, and the screening mechanism is gradually initiated in the middle and late stages of growth to avoid data distortion caused by over-screening of normal samples in the early growth stage.

[0042] S302, Dual-track Predictive Value Calculation. Based on the pre-stored standard growth curve of yellow-feathered capons, interpolation is performed using the current age at which the theoretical weight is calculated for the current growth stage, serving as a preliminary estimate. This preliminary estimate reflects the breed's inherent growth pattern and is unaffected by daily sample fluctuations, serving as one of the benchmark references for sample selection.

[0043] Based on accumulated historical average weight data, the algorithmic average weight data for the current pen over the past 3 days is selected, and a moving average algorithm is used for time-series smoothing to obtain the time-series forecast for the day. The moving average window can be adjusted according to data stability; when data fluctuations are large, it can be expanded to include data from the past 5 days to improve the smoothing effect. The time-series forecast reflects the flock's own growth trend and recent weight level, closely matching the actual feeding conditions in the pen. The dual-track forecast provides references from two dimensions: breed growth patterns and individual time-series trends, complementing each other and improving the reliability of the forecast benchmark.

[0044] S303, Determination of Dynamic Floating Ratio and Sample Screening. The baseline average weight is calculated by assigning corresponding weights to the preliminary and time-series estimated values. The weight allocation is dynamically adjusted based on the age of the chickens: the time-series estimated value has a higher weight in the early growth stage because the flock's growth is more closely related to recent feeding management; the weight of the standard growth curve gradually increases in the longer growth stage because the flock's growth tends to conform to breed patterns at this time, and individual differences narrow.

[0045] The baseline value for the floating ratio is determined based on the current growth stage of the animals' age. The growth stage is divided into four phases: brooding period, growth period, fattening period, and pre-market weight loss. The fattening period and pre-market weight loss are longer growth stages, and their corresponding baseline floating ratio values ​​are smaller. Using the baseline average weight as the center, the upper and lower limits for sample selection are calculated according to the floating ratio. The upper limit is the baseline average weight multiplied by (1 + floating ratio), and the lower limit is the baseline average weight multiplied by (1 - floating ratio).

[0046] When determining the floating ratio, first match it to the baseline ratio corresponding to the current pen's environmental type: in environmentally controlled chicken houses, the environment is stable and the weight dispersion of the chickens is low, so the baseline ratio is set at 8% to 10%; in non-environmentally controlled chicken houses, the environment fluctuates greatly, so the baseline ratio is set at 12% to 15%. Then, adjust the baseline ratio based on the current growth stage of the chickens' age and the coefficient of variation of the effective samples for the day: the higher the age of the chickens, the smaller the floating ratio, adapting to the natural law of slower weight gain and narrower fluctuations in the later stages of growth; the larger the coefficient of variation of the samples for the day, the higher the sample dispersion, so the floating ratio should be appropriately increased to avoid over-screening and removing effective samples. By dynamically adjusting the floating ratio based on multiple factors, a reasonable screening intensity can be maintained across different growth stages and environments, balancing the rigor of screening with the inclusiveness of the samples.

[0047] Using the calculated upper and lower limits, outlier samples that significantly deviate from the reasonable range are filtered from the daily valid samples, retaining valid data within the upper and lower limits. The average weight is then recalculated based on the filtered samples. Outlier samples removed during the screening process are marked, and their weight values ​​and reasons for removal are recorded for subsequent data statistics and algorithm optimization.

[0048] S400 assesses the confidence level of the calculated average weight based on the reasonableness of the daily average weight gain obtained from the algorithm's average weight over multiple consecutive days, and the variation range of the coefficient of variation obtained from the statistics of effective samples.

[0049] In this embodiment of the invention, the specific execution process of S400 is as follows: S401, a two-dimensional confidence level classification. The average daily weight gain is calculated by subtracting the average weight of the previous three days from the current day's average weight and then dividing by 3. This average daily weight gain is then compared to the reasonable weight gain range for the corresponding age of the animal. The reasonable range is based on standard growth curves and historical data, with different reasonable ranges corresponding to different growth stages. Simultaneously, the coefficient of variation (COP) of the effective sample for the day is calculated, which is the sample standard deviation divided by the average weight, reflecting the sample dispersion. The COP is then compared to the average COP of the past seven days to obtain the magnitude of the COP variation.

[0050] The daily average weight gain range and the variation range of the coefficient of variation were verified separately. Based on the results of these two verifications, the algorithm's average weight was divided into four confidence levels: excellent, good, average, and poor. When the daily average weight gain is within a reasonable range and the variation range of the coefficient of variation does not exceed the preset threshold, it is judged as excellent confidence level; when one of the two indicators is close to the critical value, or slightly exceeds it but is still within an acceptable range, it is judged as good; when one of them is significantly beyond the reasonable range, it is judged as average; when both are significantly beyond the range, or one of them deviates severely, it is judged as poor.

[0051] When the confidence level is lower than the preset level, a corresponding anomaly identifier is output and a corresponding level of review reminder is triggered: a general level triggers a regular reminder, which is pushed to the poultry administrator's terminal to prompt attention; a poor level triggers a key review reminder, and at the same time pushes the details of the abnormal data, prompting the management personnel to check the equipment status and flock condition. Through graded confidence level assessment and differentiated reminders, timely warnings are ensured when data anomalies occur, while frequent alarms are avoided from interfering with normal management.

[0052] S402, Low Confidence Triggers Supplementary Sampling and Secondary Correction. When the confidence level is lower than the preset level, and the cause of the anomaly is determined to be due to skewed sample distribution or insufficient sample representativeness, supplementary manual sampling is conducted at night in the dark for the corresponding enclosure. Supplementary sampling focuses on covering the high-weight range, with the high-weight range accounting for no less than 60% of the total supplementary sampling, specifically addressing the deficiency in high-weight samples.

[0053] Using the supplementary sampling samples and the original manually sampled samples, the baseline proportion of each weight interval under the corresponding environment is updated, and the weight coefficients of each equal interval are recalculated. Based on the updated weight coefficients, the daily valid samples are weighted again to obtain the algorithm's average weight after secondary correction, which is then output. The secondary correction process fully records the average weight values ​​before and after correction, the supplementary sampling sample size, and the adjustment range of the weight coefficients, and stores this information in the historical database for subsequent algorithm iterations and performance evaluation.

[0054] When the confidence level is abnormal due to equipment failure, flock anomalies, or other reasons, the corresponding troubleshooting process is triggered without secondary correction to avoid erroneous data misleading feeding decisions. Through supplementary sampling and secondary correction mechanisms, a closed-loop optimization of data quality is formed, effectively improving the average weight accuracy in low-confidence scenarios without increasing daily workload.

[0055] This method retains the core advantage of intelligent chicken scales—24 / 7 stress-free automatic weighing—eliminating the need for frequent manual handling that disrupts the flock's normal feeding and rest. Only a small number of samples are calibrated at night when the flock is naturally quiet, minimizing the stress impact of manual operation. The optimized structure of the extended weighing pan physically increases the rate of larger chickens actively weighing themselves, increasing the number and representativeness of daily effective samples from the outset. Combined with stress-free manual sampling under dark conditions to obtain a baseline of the flock's true weight distribution, this method solidifies the data foundation for accurate average weight calculations without disrupting the stress-free farming model.

[0056] By comparing the distribution across ten zones and dynamically adjusting the weights, this method specifically compensates for the missing samples in the high-weight zone, correcting the average weight error caused by sample skew at the distribution level. Compared to conventional global mean correction or simple outlier removal, it more accurately reflects the weight distribution characteristics of yellow-feathered capons, especially suitable for sampling large-sized, low-activity flocks, effectively improving the accuracy of average weight calculation. Simultaneously, the interval weight benchmark is dynamically updated periodically to match the distribution changes at different growth stages of the flock, maintaining consistent correction accuracy.

[0057] To address the industry pain points of unclear peak values ​​and data dispersion in the long-life, high-weight stage, this method combines standard growth curves with historical time-series data to form a two-dimensional joint prediction. This defines a reasonable sample screening boundary, maximizing the retention of effective samples while eliminating outliers, thus avoiding the smoothing distortion caused by simple mean filtering. Simultaneously, the fluctuation ratio dynamically adjusts based on multiple factors such as age, environmental type, and sample dispersion, maintaining an appropriate screening intensity across different growth stages and farming scenarios, thereby ensuring the stability of average weight data in the long-life stage from a mechanistic perspective.

[0058] Taking the fattening pen of a large-scale yellow-feathered capon farm as an example, the farm has about 1,200 birds per pen, with a breeding age of about 130 days and an estimated average weight of about 3.2 kg at slaughter. Before the upgrade, conventional smart chicken scales and simple mean calculation methods were used. The effective sample size of a single scale was about 110 to 130 birds per day, and the proportion of samples in the high weight range was only about 60% of the true distribution. The deviation between the calculated average weight and the actual average weight sold was generally between 120g and 180g, and the number of days with the deviation controlled within ±100g was only about 72%. Moreover, the data fluctuated greatly in the week before slaughter, with the daily average weight gain fluctuating by more than 60g and the weekly variation of the coefficient of variation approaching 15%. Managers could not accurately determine the slaughter time based on the data.

[0059] After applying this method, the first step was to replace the scales with smart chicken scales featuring extended weighing pans, increasing the daily effective sample size per unit to over 230 chickens and significantly expanding the sample coverage. Combined with manual sampling in the dark every three days to establish a distribution baseline, and after interval weighting correction, the average weight deviation was significantly reduced. During the long-term pre-slaughter stage, a joint prediction and sample screening mechanism was automatically triggered, with the fluctuation ratio gradually narrowing as the chickens aged, effectively filtering out abnormal data fluctuations. Final operational data showed that the deviation between the algorithm's average weight and the actual average selling weight was controlled within 45g, and the percentage of days with a deviation within ±100g increased to 93%. The average daily weight gain remained stable within a reasonable range of 15g to 35g for a consecutive week, with the coefficient of variation fluctuating by less than 8% weekly, significantly improving data stability. Simultaneously, only a small amount of sampling was required at night, without increasing stress on the flock or significantly increasing manual workload. While maintaining the advantages of stress-free farming, it simultaneously achieved improved average weight accuracy and data stability during the long-term pre-slaughter stage, providing reliable data support for refined feeding management and precise slaughter decisions.

[0060] This application discloses a system for weight processing in large-scale poultry flocks of yellow-feathered capons, referring to... Figure 2 ,include: The sample acquisition module 001 acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken in the flock via Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than the preset threshold. The standard growth curve of yellow-feathered capons is pre-stored and the historical average weight data corresponding to each feeding age is accumulated and stored. The interval weighted correction module 002 divides the daily valid samples and manually sampled samples into multiple equal intervals according to weight, calculates the sample proportion in each equal interval and compares the difference between the two proportions, determines the weight coefficient of each equal interval based on the proportion difference, and uses the weight coefficient to perform weighted correction on the daily valid samples to obtain the algorithm average weight. The sample joint screening module 003 performs joint prediction based on the pre-stored standard growth curve and the accumulated historical average weight data when the daily effective samples meet the condition of no obvious peak. After screening the daily effective samples according to the joint prediction results, the average weight is recalculated. The confidence assessment module 004 assesses the confidence of the calculated average weight based on the reasonableness of the daily average weight gain obtained from the algorithm average weight over multiple consecutive days, and the variation range of the coefficient of variation obtained from the statistics of effective samples.

[0061] This application also discloses an electronic device, including a processor, wherein the processor runs a program for the above-described method for processing the average weight of a large-scale poultry flock of yellow-feathered capons.

[0062] This application also discloses a storage medium storing a program for the above-described method of treating the average weight of a large-scale chicken flock for yellow-feathered capons.

[0063] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for processing a flock of broiler breeders for uniformity in weight in a large scale farming of yellow feathered breeders, characterized in that, include: The system acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken individually using the Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than the preset threshold. The system also pre-stores the standard growth curve of yellow-feathered capons and accumulates and stores the historical average weight data corresponding to each feeding age. The daily valid samples and manually sampled samples are divided into multiple equal intervals according to weight. The sample proportion in each equal interval is calculated and the difference between the two proportions is compared. The weight coefficient of each equal interval is determined based on the difference in proportion. The daily valid samples are then weighted and corrected using the weight coefficient to obtain the algorithm average weight. When the daily effective sample meets the condition that the peak is not obvious, a joint prediction is made based on the pre-stored standard growth curve and the accumulated historical average weight data. The average weight is then recalculated after screening the daily effective samples according to the joint prediction results. The confidence level of the calculated average weight is assessed based on the reasonableness of the daily average weight gain obtained from the algorithmic average weight over several consecutive days, and the variation range of the coefficient of variation obtained from the statistics of effective samples.

2. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 1, characterized in that, The steps for obtaining daily valid samples continuously collected by the smart chicken scale and manually sampled samples include: A smart chicken scale with an extended weighing pan is used to collect data continuously throughout the day. The collected data is processed to remove jitter, identify multiple chickens, and remove outliers to obtain a daily valid sample. In the dark at night, the flock of chickens is delineated centered on the location where the smart chicken scale is placed. Each chicken is weighed individually using the Bluetooth chicken scale to obtain a manually sampled sample. The manually sampled sample covers three weight ranges: small weight, medium weight, and large weight.

3. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 1, characterized in that, The steps for dividing daily valid samples and manually sampled samples into multiple equal intervals according to weight and then weighting and adjusting them to obtain the algorithm's average weight include: The daily valid samples and manually sampled samples were divided into multiple equal intervals according to weight from smallest to largest, and the sample proportion in each interval was calculated. By comparing the differences in the proportion of the two types of samples in the corresponding intervals, the weight coefficient is increased for the high-weight intervals in the daily valid samples that have a lower proportion than the manually sampled samples, and the weight coefficient is decreased for the intervals that have a higher proportion than the manually sampled samples. After normalizing the weight coefficients of each interval, the average weight of the algorithm is obtained by weighting the effective samples of each day.

4. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 3, characterized in that, The steps for joint prediction and recalculation of the average weight when the daily effective sample meets the condition of no significant peak include: When the peak of the weight distribution of the daily effective samples is not obvious, a preliminary estimate is obtained based on the pre-stored standard growth curve of yellow-feathered capons and the current age of the chickens. Time series prediction values ​​are obtained by performing time series smoothing based on accumulated historical average weight data. Based on the preliminary and time-series estimates, the upper and lower limits of sample selection are determined. After filtering out abnormal samples from the daily valid samples, the average weight is recalculated.

5. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 1, characterized in that, The process of manually sampling chickens by weighing them one by one using a Bluetooth chicken scale includes: Representative pens were manually sampled in both environmentally controlled and non-environmentally controlled chicken houses, and weighing data were recorded under different conditions. If the number of samples in a single manual sampling exceeds a preset threshold, manual sampling will be repeated at fixed intervals to update the baseline proportion of each weight range under the corresponding environment.

6. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 4, characterized in that, The steps for determining the upper and lower limits of sample selection based on preliminary and time-series estimates include: The baseline average weight value is calculated by assigning corresponding weights to the preliminary estimate and the time series estimate, respectively. The floating ratio is determined based on the current growth stage of the animals at their current age, and the upper and lower limits of sample selection are calculated based on the baseline average weight. The longer the animal is kept, the smaller the corresponding fluctuation ratio.

7. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 1, characterized in that, The steps for evaluating the confidence level of the calculated algorithm average weight include: The daily average weight gain calculated from the algorithm's average weight over several consecutive days and the variation range of the coefficient of variation obtained from the statistical analysis of the effective samples on the day are verified separately. Based on the results of the two verifications, the algorithm's average weight is divided into different confidence levels. When the confidence level is lower than the preset level, the corresponding abnormality flag is output and the corresponding level of review reminder is triggered.

8. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 7, characterized in that, include: When the confidence level is lower than the preset level, supplementary manual sampling is carried out for the corresponding enclosure under dark conditions at night. The sampling focuses on covering the higher weight range. The samples obtained from the supplementary sampling are used to update the weight range ratio benchmark under the corresponding environment, and the weight coefficient of each equal interval is re-determined. The average weight of the algorithm for the day is then corrected a second time before being output.

9. The method for treating the average weight of large-scale poultry flocks of yellow-feathered capons according to claim 6, characterized in that, include: When determining the floating ratio, first match the base ratio to the environmental type corresponding to the current pen, and then adjust the base ratio based on the growth stage of the current age of the animals and the coefficient of variation of the effective samples on the day. The higher the age of the animals, the smaller the floating ratio, and the larger the coefficient of variation, the larger the floating ratio.

10. A system for weight processing in large-scale poultry flocks of yellow-feathered capons, characterized in that, include: The sample acquisition module acquires daily valid samples continuously collected by the smart chicken scale, as well as manually sampled samples obtained by weighing each chicken in the flock via Bluetooth chicken scale under dark conditions at night. The number of manually sampled samples is greater than the preset threshold. The standard growth curve of yellow-feathered capons is pre-stored and the historical average weight data corresponding to each feeding age is accumulated and stored. The interval weighted correction module divides the daily valid samples and manually sampled samples into multiple equal intervals according to weight, calculates the sample proportion in each equal interval and compares the difference between the two proportions, determines the weight coefficient of each equal interval based on the proportion difference, and uses the weight coefficient to perform weighted correction on the daily valid samples to obtain the algorithm average weight. The sample joint screening module performs joint prediction based on the pre-stored standard growth curve and the accumulated historical average weight data when the daily effective samples meet the condition of no obvious peak. After screening the daily effective samples according to the joint prediction results, the average weight is recalculated. The confidence assessment module assesses the confidence of the calculated average weight based on the reasonableness of the daily average weight gain obtained from the algorithm's average weight over multiple consecutive days, and the variation range of the coefficient of variation obtained from the statistics of valid samples.