Method for detecting abnormal poultry feed intake and breeding feed additive

Abstract

This invention relates to the field of poultry farming technology, and more particularly to a method for detecting abnormal feed intake in poultry and a feed additive for poultry farming. The method includes: acquiring continuous images of a target poultry flock within a monitoring period and preprocessing the continuous images; identifying the feeding behavior of individual target poultry based on the continuous images of the target poultry flock, and then constructing the basic behavioral characteristics of the target individuals; extracting the physiological state characteristics of the target individuals based on the continuous images of the target poultry flock; acquiring environmental parameters in real time and constructing an environmental anomaly index based on the environmental parameters, and then establishing a feeding anomaly judgment model for the target individuals by combining the basic behavioral characteristics and physiological state characteristics of the target individuals, and outputting the feeding judgment result; constructing a regional anomaly index for the target poultry flock based on the feeding status of the target individuals, and then issuing an anomaly alarm to the user. This invention effectively improves the efficiency and accuracy of detecting abnormal feed intake in poultry.

Description

Technical Field

[0001] This invention relates to the field of poultry farming technology, and in particular to a method for detecting abnormal feed intake in poultry and a feed additive for poultry farming. Background Technology

[0002] In modern large-scale poultry farming, feed intake is a key indicator reflecting the health status and growth performance of poultry. Accurate monitoring of feeding behavior is of great significance for disease prevention, optimization of feed formulation, and improvement of farming efficiency. However, in traditional farming models, it is difficult for farmers to monitor the individual feeding status of a large number of poultry in real time, especially in stacked cage environments, where manual observation is inefficient and easily affected by subjective factors.

[0003] Currently, the industry generally uses manual statistics of daily feed intake for anomaly detection. This method cannot distinguish individual differences and is difficult to separate the combined effects of factors such as age and environmental fluctuations on feed intake. Static daily data comparison cannot capture early, latent abnormal signals and lacks the ability to accurately trace the cause of anomalies, often leading to delayed intervention measures, feed waste, and decreased growth performance. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting abnormal feed intake in poultry and a feed additive for livestock, so as to solve at least one of the problems existing in the prior art.

[0005] To achieve the above objectives, according to one aspect of this application, the present invention provides a method for detecting abnormal feed intake in poultry, comprising:

[0006] Collect continuous images of the target poultry flock during the monitoring period and preprocess the continuous images;

[0007] Based on continuous images of the target poultry flock, the feeding behavior of individual target individuals is identified, thereby constructing the basic behavioral characteristics of the target individuals;

[0008] Physiological characteristics of individual target poultry are extracted from continuous images of the target poultry flock;

[0009] Real-time collection of environmental parameters, construction of an environmental anomaly index based on environmental parameters, and then establishment of a feeding anomaly judgment model for the target individual by combining the basic behavioral characteristics and physiological state characteristics of the target individual, and output of feeding judgment results;

[0010] A regional anomaly index for the target poultry flock is constructed based on the feeding status of the target individuals within the management cycle, and then anomaly alerts are sent to the user.

[0011] Optionally, a high-definition network camera is installed directly above each row of cages to capture the activities of the target individuals inside the cages from above, so as to collect continuous images. In this application, the frame rate for capturing continuous images is 30fps.

[0012] Preprocessing of continuous images:

[0013] An adaptive median filtering algorithm is used to remove image noise, with the filter window size set to 3×3 pixels.

[0014] Image contrast is enhanced by histogram equalization, and the processed video stream is divided and stored in units according to the monitoring period, with each frame of the image being timestamped.

[0015] Optionally, behavior recognition is performed on continuous images within the monitoring period, and the recognition results are statistically analyzed to obtain feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i). The statistical results are then listed as the basic behavioral characteristics of the i-th target individual, where i is the numerical ID of the target individual.

[0016] Optionally, based on continuous images of the target poultry flock, the body extension index and individual activity index of the target individuals are constructed, and the body extension index and individual activity index of the target individuals are used to construct the physiological state characteristics of the target individuals:

[0017] The postural extension index of the target individual is set as the ratio of the target individual's trunk length to trunk width;

[0018] The formula for calculating the individual activity index of the target individual is as follows: In the formula, α(i) represents the individual activity index of the i-th target individual, and d(i,j,k) represents the distance moved between the j-1-th image and the j-th key bone point in the j-th image of the i-th target individual during the monitoring period.

[0019] Optionally, the body extension index and individual activity index of the target individual are subjected to minimum-maximum normalization, and then the physiological state characteristic β(i) of the target individual is calculated based on the normalized body extension index and individual activity index: set β(i) = a1 × body extension index of the i-th target individual + a2 × individual activity index of the i-th target individual; a1 and a2 are the extension weight and activity weight, respectively.

[0020] Optionally, environmental parameters can be collected in real time and an environmental anomaly index can be constructed to correct the physiological state characteristics of the target individual.

[0021] The feeding abnormality coefficient of the target individual is established based on the physiological state characteristics and basic behavioral characteristics of the target individual after correction;

[0022] The feeding status of the target individual is determined based on the feeding abnormality index.

[0023] Optionally, an environmental anomaly index EI is constructed based on environmental parameters, and EI is set as 1−[b1×(THI−THI0) / THI0+b2×(CNH3−C0) / C0+b3×|L−L0| / L0+b4×(V0−V) / V0];

[0024] In the formula, THI is the daily average temperature and humidity index of the chicken house, CNH3 is the daily average ammonia concentration, L is the real-time light intensity, V is the hourly ventilation volume, THI0 is the suitable temperature and humidity index threshold for the target poultry breed, C0 is the upper limit threshold for ammonia concentration in poultry farming, L0 is the suitable light intensity threshold for poultry, and V0 is the suitable ventilation volume threshold for poultry farming. The above thresholds are all set according to the farming industry standards for different poultry breeds, and b1, b2, b3, and b4 are all weighting coefficients.

[0025] The product of the environmental anomaly index EI and the physiological state characteristic β(i) is used as the corrected physiological state characteristic of the target individual.

[0026] Optionally, the basic behavioral characteristics of the target individual, namely feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i), are subjected to min-max normalization.

[0027] Based on the normalized basic behavioral characteristics, a comprehensive basic behavioral characteristic F(i) of the target individual is constructed, and the calculation formula is set as: F(i)=c1×f'(i)+c2×t'(i)+c3×ft'(i); where f'(i), t'(i), and ft'(i) are the feeding frequency, average feeding duration, and average feeding interval after minimum-maximum normalization, respectively, and c1, c2, and c3 are the weight coefficients of the basic behavioral characteristics.

[0028] The construction of the feeding abnormality coefficient involves weighted fusion of the corrected physiological state characteristics and the basic behavioral comprehensive characteristics F(i) to construct the feeding abnormality coefficient ε(i) of the target individual. The calculation formula is set as: ε(i)=d1×F(i)+d2×β'(i); where β'(i) is the corrected physiological state characteristics of the i-th target individual, and d1 and d2 are the fusion weight coefficients.

[0029] Optionally, various abnormal thresholds are set, and the feeding abnormality index of the target individual is compared with each abnormal threshold to determine the feeding status of the target individual:

[0030] When ε(i) is greater than or equal to 0.7, the feeding status of the i-th target individual is determined to be without risk of abnormal feeding amount.

[0031] When ε(i) is less than 0.7, the feeding status of the i-th target individual is determined to be abnormal feeding.

[0032] A regional anomaly index R is constructed for each monitoring area based on the feeding status of target individuals within the management period, and R is set as 1 / K×Σ[w(k)×P(k)]; where K is the number of target individuals in the monitoring area, and w(k) represents the anomaly weight coefficient of the k-th target individual in the monitoring area, and w(k) is set as 1×μ n(k) n(k) represents the number of times the k-th target individual exhibits severe feeding abnormalities within the management period, μ is the baseline multiplication factor, and P(k) represents the average value of the feeding abnormality coefficient of the k-th target individual within the management period.

[0033] When R×e1+h×e2 is greater than 0.7, the monitoring area is considered normal and no alarm is triggered to the user. In the formula, e1 and e2 are regional weights, and h is the proportion of target individuals with abnormal feed intake during the management period.

[0034] When R×e1+h×e2 is less than or equal to 0.7, it is determined that there is a group risk in the monitored area, and an area risk alarm is sent to the user.

[0035] According to another aspect of this application, a livestock feed additive is also provided, comprising:

[0036] The content of γ-aminobutyric acid (GABA), vitamin C, and bentonite in animal feed additives is as follows: GABA accounts for 80%, vitamin C accounts for 2%, and bentonite accounts for 18%.

[0037] Compared with existing technologies, the beneficial effects of this invention are as follows: This solution, through the system integration of multimodal sensing, dynamic modeling, and intelligent analysis technologies, constructs a full-process intelligent system covering feed intake monitoring, anomaly early warning, source tracing diagnosis, and management push notifications. It not only achieves seamless and accurate monitoring of individual feeding behavior and early risk warnings (with the warning window shifted forward by 12-24 hours) in stacked cage rearing environments, but also significantly improves the scientific nature and response efficiency of management decisions based on data correlation analysis and hierarchical push mechanisms. Overall, it effectively reduces labor costs and the risk of misjudgment, minimizes feed waste and disease transmission losses, promotes the shift from experience-driven to data-driven livestock management, comprehensively improves breeding efficiency, animal health, and the level of industry intelligence, and provides practical technical support for modern smart animal husbandry. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the method for detecting abnormal feed intake in poultry in this embodiment.

[0040] Figure 2 This is a flowchart illustrating the abnormal feeding detection method in this embodiment. Detailed Implementation

[0041] To more clearly illustrate the present invention, the following description, in conjunction with preferred embodiments and accompanying drawings, further clarifies the invention. Similar components in the drawings are indicated by the same reference numerals. Those skilled in the art should understand that the specific description below is illustrative rather than restrictive and should not be construed as limiting the scope of protection of the present invention.

[0042] It should be noted that although the terms first, second, third, etc., may be used in the embodiments of this application for description, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, without departing from the scope of the embodiments of this application, first can also be referred to as second, and similarly, second can also be referred to as first.

[0043] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0044] Specifically, the poultry feed intake abnormality detection method and feed additives described in this application are applied to the detection of abnormal feed intake in large-scale caged broiler chickens aged 1-42 days from brooding to growing period; the large-scale caged broiler chickens described in this application are specifically raised in a stacked chicken house.

[0045] Specifically, in this application scenario, broiler chickens are prone to abnormal feed intake due to rapid growth, fluctuations in the breeding environment, differences in group feeding behavior, and early latent health problems. Existing detection technologies mostly rely on manual statistics of total feed consumption to make static daily value judgments, which cannot separate the data interference caused by the natural growth pattern of age, stress in the breeding environment, and equipment feeding deviations. It is difficult to capture early latent signals of abnormal feed intake, and it is impossible to accurately trace the type of abnormality and the core cause.

[0046] To apply to the above-mentioned application scenarios, this application provides a method for detecting abnormal feed intake in poultry, the flowchart of which can be found in the document. Figure 1 As shown, it includes:

[0047] Step S100: Collect continuous images of the target poultry flock during the monitoring period and preprocess the continuous images.

[0048] Specifically, the implementation process of step S100 is as follows:

[0049] A high-definition network camera is installed directly above each row of cages to capture the activities of the target individuals inside the cages from above, so as to collect continuous images. The frame rate for capturing continuous images in this application is 30fps.

[0050] Preprocessing of continuous images:

[0051] An adaptive median filtering algorithm is used to remove image noise, with the filter window size set to 3×3 pixels.

[0052] Image contrast is enhanced by histogram equalization, and the processed video stream is divided and stored in units according to the monitoring period, with each frame of the image being timestamped.

[0053] Specifically, this application does not impose a specific limitation on the duration of the monitoring period. Those skilled in the art can set it freely, as long as the value requirement of the monitoring period duration is met. In this application, the duration of the monitoring period is set to 30 minutes.

[0054] Specifically, a non-contact multimodal sensor array is deployed to achieve unobtrusive and continuous monitoring of poultry feeding behavior in stacked cage environments. This step effectively avoids blind spots and subjective biases in manual observation, collecting micro-parameters such as individual feeding time, frequency, and amount of food with second-level accuracy to build a high-fidelity behavioral database. At the same time, it avoids animal stress reactions, significantly improving data objectivity and collection efficiency, laying a solid foundation for accurate analysis.

[0055] Please continue reading. Figure 1 As shown, the method for detecting abnormal feed intake in poultry also includes:

[0056] Step S200: Based on the continuous images of the target poultry flock, identify the feeding behavior of the target individuals, and then construct the basic behavioral characteristics of the target individuals.

[0057] Specifically, before identifying the feeding behavior of a target individual based on continuous images of the target poultry flock in step S200, the following steps need to be completed:

[0058] The YOLOv8-Pose model is used to detect target individuals and identify key points in each preprocessed frame of image. The detection confidence threshold of the YOLOv8-Pose model is set to 0.85, and the key point confidence threshold is set to 0.7. Continuous images are input into the model, and the model outputs the instance bounding box of each target individual and the two-dimensional coordinates of 14 key skeletal points. A unique identity ID is assigned to each target individual. The key skeletal points mentioned in this application include: beak tip, head center, eyes, upper cervical joint, lower cervical joint, anterior trunk, posterior trunk, left wing root, right wing root, left leg root, right leg root, left toe, right toe, and tail tip.

[0059] Mark the trough area in the continuous image.

[0060] Specifically, the marking of the trough area in this application

[0061] The implementation process of step S200 is as follows:

[0062] Behavior recognition is performed on continuous images within the monitoring period, and the recognition results are statistically analyzed to obtain feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i). The statistical results are then listed as the basic behavioral characteristics of the i-th target individual, where i is the numerical ID of the target individual.

[0063] For example, in this application, when performing behavior recognition on continuous images within a monitoring period, the pre-collected continuous images of poultry feeding behavior are input into a three-dimensional convolutional network to extract feeding behavior features, and the continuous images within the monitoring period are input into the three-dimensional convolutional network to identify the number of feeding behaviors contained therein.

[0064] Specifically, in this application, feeding behavior is defined as follows: the target individual's beak tip bone point coordinates enter the feed trough area, the head movement speed is greater than 0.3 m / s, the head retraction action occurs within 0.1 s after the beak tip bone point enters the feed trough area, and the retraction speed exceeds 0.2 m / s; a complete feeding behavior is determined when all of the above conditions are met; the average feeding duration is defined as the average duration of the feeding behavior within the monitoring period, and the average feeding interval duration is defined similarly.

[0065] Specifically, an adaptive benchmark model integrating dynamic parameters such as age, weight, and environmental temperature and humidity is constructed to replace the traditional static threshold judgment. This step enables the system to intelligently calibrate the normal feeding range according to the breeding cycle, effectively eliminate external interference factors, significantly improve the accuracy and robustness of anomaly identification, reduce false alarms and missed alarms, make the monitoring logic more in line with the physiological laws of poultry flocks, and enhance the practicality and reliability of the system.

[0066] Please continue reading. Figure 1 As shown, the method for detecting abnormal feed intake in poultry also includes:

[0067] Step S300: Extract physiological state characteristics of target individuals based on continuous images of the target poultry flock.

[0068] Specifically, the implementation process of step S300 is as follows:

[0069] Based on continuous images of the target poultry flock, the body extension index and individual activity index of the target individuals were constructed, and the body extension index and individual activity index of the target individuals were used to construct the physiological state characteristics of the target individuals:

[0070] The postural extension index of the target individual is set as the ratio of the target individual's trunk length to trunk width;

[0071] The formula for calculating the individual activity index of the target individual is as follows: In the formula, α(i) represents the individual activity index of the i-th target individual, and d(i,j,k) represents the distance moved between the j-1-th image and the j-th key bone point in the j-th image of the i-th target individual during the monitoring period.

[0072] The body extension index and individual activity index of the target individual are min-max normalized, and then the physiological state characteristics β(i) of the target individual are calculated based on the normalized body extension index and individual activity index: set β(i) = a1 × body extension index of the i-th target individual + a2 × individual activity index of the i-th target individual; a1 and a2 are the extension weight and activity weight, respectively.

[0073] Specifically, the trunk length mentioned in this application is the Euclidean distance from the front end to the rear end of the trunk, and the trunk width is the Euclidean distance between the roots of the left and right wings.

[0074] Specifically, a sliding window time-series analysis algorithm is introduced to perform real-time pattern recognition on continuous data streams, sensitively capturing early, weak abnormal signals such as a sudden drop in feeding frequency. Compared to daily total comparison, this step can advance the risk warning time by 12-24 hours, achieving a key shift from "post-event statistics" to "pre-event intervention," securing a golden window for disease prevention and management decisions, and significantly reducing the risk of losses in aquaculture.

[0075] Please continue reading. Figure 1 As shown, the method for detecting abnormal feed intake in poultry also includes:

[0076] Step S400: Collect environmental parameters in real time, construct an environmental anomaly index based on the environmental parameters, and then combine the basic behavioral characteristics and physiological state characteristics of the target individual to establish a feeding anomaly judgment model for the target individual and output the feeding judgment result.

[0077] Please see Figure 2 The diagram shown is a flowchart of the feeding abnormality detection method provided in this application, including:

[0078] Step S401: Collect environmental parameters in real time and construct an environmental anomaly index to correct the physiological state characteristics of the target individual.

[0079] Step S402: Establish the feeding abnormality coefficient of the target individual based on the physiological state characteristics and basic behavioral characteristics of the target individual after correction.

[0080] Step S403: Determine the feeding status of the target individual based on the feeding abnormality index of the target individual.

[0081] Specifically, the implementation process of step S401 is as follows:

[0082] An environmental anomaly index EI is constructed based on environmental parameters, and EI is set as 1−[b1×(THI−THI0) / THI0+b2×(CNH3−C0) / C0+b3×|L−L0| / L0+b4×(V0−V) / V0];

[0083] In the formula, THI is the daily average temperature and humidity index of the chicken house, CNH3 is the daily average ammonia concentration, L is the real-time light intensity, V is the hourly ventilation volume, THI0 is the suitable temperature and humidity index threshold for the target poultry breed, C0 is the upper limit threshold for ammonia concentration in poultry farming, L0 is the suitable light intensity threshold for poultry, and V0 is the suitable ventilation volume threshold for poultry farming. The above thresholds are all set according to the farming industry standards for different poultry breeds, and b1, b2, b3, and b4 are all weighting coefficients.

[0084] The product of the environmental anomaly index EI and the physiological state characteristic β(i) is used as the corrected physiological state characteristic of the target individual.

[0085] It is understood that in this application, when THI is less than THI0, (THI−THI0) is 0; when CNH3 is less than C0, (CNH3−C0) is 0; when V is greater than V0, (V0−V) is 0, that is, only the abnormal part of the environmental parameters exceeding the appropriate threshold is calculated on the environmental anomaly index.

[0086] Specifically, the implementation process of step S402 is as follows:

[0087] The basic behavioral characteristics of the target individual, namely feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i), are subjected to min-max normalization.

[0088] Based on the normalized basic behavioral characteristics, a comprehensive basic behavioral characteristic F(i) of the target individual is constructed, and the calculation formula is set as: F(i)=c1×f'(i)+c2×t'(i)+c3×ft'(i); where f'(i), t'(i), and ft'(i) are the feeding frequency, average feeding duration, and average feeding interval after minimum-maximum normalization, respectively, and c1, c2, and c3 are the weight coefficients of the basic behavioral characteristics, and c1+c2+c3=1.

[0089] The construction of the feeding abnormality coefficient involves weighted fusion of the corrected physiological state characteristics and the basic behavioral comprehensive characteristics F(i) to construct the feeding abnormality coefficient ε(i) of the target individual. The calculation formula is set as: ε(i)=d1×F(i)+d2×β'(i); where β'(i) is the corrected physiological state characteristics of the i-th target individual, d1 and d2 are the fusion weight coefficients, and d1+d2=1.

[0090] Specifically, in this application, c1=0.45, c2=0.35, c3=0.2, and d1=0.6, d2=0.4 are set.

[0091] Specifically, the implementation process of step S403 is as follows:

[0092] Set various abnormal thresholds and compare the target individual's feeding abnormality index with each abnormal threshold to determine the target individual's feeding status:

[0093] When ε(i) is greater than or equal to 0.7, the feeding status of the i-th target individual is determined to be without risk of abnormal feeding amount.

[0094] When ε(i) is less than 0.7, the feeding status of the i-th target individual is determined to be abnormal feeding.

[0095] Please continue reading. Figure 1 As shown, the method for detecting abnormal feed intake during the Jiaqing reign also includes:

[0096] Step S500: Based on the feeding status of the target individuals within the management period, construct a regional abnormality index for the target poultry flock, and then issue an abnormality alarm to the user.

[0097] Specifically, before performing step S500, the stacked chicken coop described in this application needs to be divided into monitoring areas by physical location, and each monitoring area contains the same number of target individuals; in this application, an area with 8 columns and 4 rows, totaling 32 target individuals, is used as a monitoring area; the duration of the management cycle described in this application is an integer multiple of the monitoring cycle, and the multiple is greater than 30. In this application, the duration of the management cycle is 48 times the duration of the monitoring cycle.

[0098] Specifically, a multi-source data association and traceability mechanism is established. When an anomaly is triggered, information such as environmental records and group behavior within the same cage is automatically integrated for intelligent attribution analysis. This step can accurately distinguish between different causes such as disease, feed palatability, or equipment malfunction, generating structured diagnostic suggestions to help farmers quickly locate the root cause of the problem, avoid blind treatment, and improve problem-solving efficiency and management scientificity.

[0099] Specifically, the implementation process of step S500 is as follows:

[0100] A regional anomaly index R is constructed for each monitoring area based on the feeding status of target individuals within the management period, and R is set as 1 / K×Σ[w(k)×P(k)]; where K is the number of target individuals in the monitoring area, and w(k) represents the anomaly weight coefficient of the k-th target individual in the monitoring area, and w(k) is set as 1×μ n(k) n(k) represents the number of times the k-th target individual exhibits severe feeding abnormalities within the management period, μ is the baseline multiplication factor, and P(k) represents the average value of the feeding abnormality coefficient of the k-th target individual within the management period.

[0101] When R×e1+h×e2 is greater than 0.7, the monitoring area is considered normal and no alarm is triggered to the user. In the formula, e1 and e2 are regional weights, and h is the proportion of target individuals with abnormal feed intake during the management period.

[0102] When R×e1+h×e2 is less than or equal to 0.7, it is determined that there is a group risk in the monitored area, and an area risk alarm is sent to the user.

[0103] Specifically, the base multiplication factor described in this application is 1.1, e1=0.8, and e2=0.2.

[0104] In conjunction with the aforementioned method for detecting abnormal feed intake in poultry, this application also provides a livestock feed additive, comprising:

[0105] The content of γ-aminobutyric acid (GABA), vitamin C, and bentonite in animal feed additives is as follows: GABA accounts for 80%, vitamin C accounts for 2%, and bentonite accounts for 18%.

[0106] Meanwhile, the proportion of the aquaculture feed additives described in this application in the feed is 0.5%.

[0107] Specifically, a tiered early warning and push system is designed to send customized prompts and handling suggestions to managers via mobile devices based on the level of anomaly. This step achieves efficient integration of early warning information and management actions, supports historical trend comparison and decision optimization, promotes the shift from experience-driven to data-driven aquaculture management, effectively reduces feed waste, and improves flock health and overall aquaculture economic benefits.

[0108] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for detecting abnormal feed intake in poultry, characterized in that, include: Collect continuous images of the target poultry flock during the monitoring period and preprocess the continuous images; Based on continuous images of the target poultry flock, the feeding behavior of individual target individuals is identified, thereby constructing the basic behavioral characteristics of the target individuals; Physiological characteristics of individual target poultry are extracted from continuous images of the target poultry flock; Real-time collection of environmental parameters, construction of an environmental anomaly index based on environmental parameters, and then establishment of a feeding anomaly judgment model for the target individual by combining the basic behavioral characteristics and physiological state characteristics of the target individual, and output of feeding judgment results; A regional anomaly index for the target poultry flock is constructed based on the feeding status of the target individuals within the management cycle, and then anomaly alerts are sent to the user.

2. The method for detecting abnormal feed intake in poultry according to claim 1, characterized in that, A high-definition network camera is installed directly above each row of cages to capture the activities of the target individuals inside the cages from above, so as to collect continuous images. The frame rate for capturing continuous images in this application is 30fps. Preprocessing of continuous images: An adaptive median filtering algorithm is used to remove image noise, with the filter window size set to 3×3 pixels. Image contrast is enhanced by histogram equalization, and the processed video stream is divided and stored in units according to the monitoring period, with each frame of the image being timestamped.

3. The method for detecting abnormal feed intake in poultry according to claim 2, characterized in that, Behavior recognition is performed on continuous images within the monitoring period, and the recognition results are statistically analyzed to obtain feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i). The statistical results are then listed as the basic behavioral characteristics of the i-th target individual, where i is the numerical ID of the target individual.

4. The method for detecting abnormal feed intake in poultry according to claim 3, characterized in that, Based on continuous images of the target poultry flock, the body extension index and individual activity index of the target individuals were constructed, and the body extension index and individual activity index of the target individuals were used to construct the physiological state characteristics of the target individuals: The postural extension index of the target individual is set as the ratio of the target individual's trunk length to trunk width; The formula for calculating the individual activity index of the target individual is as follows: In the formula, α(i) represents the individual activity index of the i-th target individual, and d(i,j,k) represents the distance moved between the j-1-th image and the j-th key bone point in the j-th image of the i-th target individual during the monitoring period.

5. The method for detecting abnormal feed intake in poultry according to claim 4, characterized in that, The body extension index and individual activity index of the target individual are min-max normalized, and then the physiological state characteristics β(i) of the target individual are calculated based on the normalized body extension index and individual activity index: set β(i) = a1 × body extension index of the i-th target individual + a2 × individual activity index of the i-th target individual; a1 and a2 are the extension weight and activity weight, respectively.

6. The method for detecting abnormal feed intake in poultry according to claim 5, characterized in that, Real-time collection of environmental parameters and construction of an environmental anomaly index are used to correct the physiological state characteristics of the target individual. The feeding abnormality coefficient of the target individual is established based on the physiological state characteristics and basic behavioral characteristics of the target individual after correction; The feeding status of the target individual is determined based on the feeding abnormality index.

7. The method for detecting abnormal feed intake in poultry according to claim 6, characterized in that, An environmental anomaly index EI is constructed based on environmental parameters, and EI is set as 1−[b1×(THI−THI0) / THI0+b2×(CNH3−C0) / C0+b3×|L−L0| / L0+b4×(V0−V) / V0]; In the formula, THI is the daily average temperature and humidity index of the chicken house, CNH3 is the daily average ammonia concentration, L is the real-time light intensity, V is the hourly ventilation volume, THI0 is the suitable temperature and humidity index threshold for the target poultry breed, C0 is the upper limit threshold for ammonia concentration in poultry farming, L0 is the suitable light intensity threshold for poultry, and V0 is the suitable ventilation volume threshold for poultry farming. The above thresholds are all set according to the farming industry standards for different poultry breeds, and b1, b2, b3, and b4 are all weighting coefficients. The product of the environmental anomaly index EI and the physiological state characteristic β(i) is used as the corrected physiological state characteristic of the target individual.

8. The method for detecting abnormal feed intake in poultry according to claim 7, characterized in that, The basic behavioral characteristics of the target individual, namely feeding frequency f(i), average feeding duration t(i), and average feeding interval ft(i), are subjected to min-max normalization. Based on the normalized basic behavioral characteristics, a comprehensive basic behavioral characteristic F(i) of the target individual is constructed, and the calculation formula is set as: F(i) = c1 × f'(i) + c2 × t'(i) + c3 × ft'(i); where f'(i), t'(i), and ft'(i) are the feeding frequency, average feeding duration, and average feeding interval after minimum-maximum normalization, respectively, and c1, c2, and c3 are the weight coefficients of the basic behavioral characteristics; The construction of the feeding abnormality coefficient involves weighted fusion of the corrected physiological state characteristics and the basic behavioral comprehensive characteristics F(i) to construct the feeding abnormality coefficient ε(i) of the target individual. The calculation formula is set as: ε(i)=d1×F(i)+d2×β'(i); where β'(i) is the corrected physiological state characteristics of the i-th target individual, and d1 and d2 are the fusion weight coefficients.

9. The method for detecting abnormal feed intake in poultry according to claim 8, characterized in that, Set various abnormal thresholds and compare the target individual's feeding abnormality index with each abnormal threshold to determine the target individual's feeding status: When ε(i) is greater than or equal to 0.7, the feeding status of the i-th target individual is determined to be without risk of abnormal feeding amount. When ε(i) is less than 0.7, the feeding status of the i-th target individual is determined to be abnormal feeding. A regional anomaly index R is constructed for each monitoring area based on the feeding status of target individuals within the management period, and R is set as 1 / K×Σ[w(k)×P(k)]; where K is the number of target individuals in the monitoring area, and w(k) represents the anomaly weight coefficient of the k-th target individual in the monitoring area, and w(k) is set as 1×μ n(k) n(k) represents the number of times the k-th target individual exhibits severe feeding abnormalities within the management period, μ is the baseline multiplication factor, and P(k) represents the average value of the feeding abnormality coefficient of the k-th target individual within the management period. When R×e1+h×e2 is greater than 0.7, the monitoring area is considered normal and no alarm is triggered to the user. In the formula, e1 and e2 are regional weights, and h is the proportion of target individuals with abnormal feed intake during the management period. When R×e1+h×e2 is less than or equal to 0.7, it is determined that there is a group risk in the monitored area, and an area risk alarm is sent to the user.

10. A livestock feed additive, comprising: The content of γ-aminobutyric acid (GABA), vitamin C, and bentonite in animal feed additives is as follows: GABA accounts for 80%, vitamin C accounts for 2%, and bentonite accounts for 18%.

Images