A poultry feed intake measurement device and method

Abstract

The present application discloses a kind of poultry intake determination device and method, method includes: the three-order nonlinear static calibration model of weighing sensor and temperature drift compensation calibration model are constructed;Combining binocular vision recognition and laser ranging technology determines the effective state of feeding process, obtains the effective feeding period of poultry;Based on the time corresponding to feeding point, the weight value collected by weighing sensor is obtained, and the intake data of poultry in effective feeding period is obtained;According to intake data, the total intake and feeding efficiency of poultry in effective feeding period are calculated, and the time label of each effective feeding period is obtained.The device includes: the lower end of L-shaped support rod is connected with weighing sensor, and the lower end of weighing sensor is provided with a hopper through L-shaped hanger;The upper end of L-shaped support rod is provided with binocular vision recognition module, laser ranging sensor and infrared temperature sensor;It also includes a data management system arranged in the breeding field control center and an RFID module arranged on the breeding cage.

Description

Technical Field

[0001] This invention relates to the field of poultry farming, and specifically to a device and method for measuring poultry feed intake. Background Technology

[0002] Poultry feed intake is a core indicator for evaluating poultry production performance, feed nutritional value, breeding superior breeds, and monitoring poultry health status. Accurate feed intake measurement is fundamental to poultry farming, breeding, and nutritional research. Currently, feed intake in poultry farming is mostly recorded manually, calculated by periodically weighing the initial and remaining feed in the feed trough. This only provides static cumulative values ​​for the entire group or at intervals, failing to provide real-time dynamic monitoring. Human operation is prone to error and can cause stress to poultry, making it unsuitable for large-scale farming scenarios. Furthermore, current technologies only output static cumulative feed intake, failing to obtain refined dynamic parameters such as instantaneous feed intake per pecking, feeding efficiency, and feeding frequency, thus hindering data support for poultry feeding behavior and nutritional metabolism research.

[0003] In modern poultry farming, feed intake is a core indicator for measuring poultry production performance, assessing feed nutritional value, selecting superior breeds, and monitoring health status. Accurate and real-time feed intake measurement is not only the foundation of scientific farming management but also a crucial link in achieving precise nutrient supply, reducing feed costs, and improving farming efficiency. However, current farming practices still primarily rely on manually weighing feed troughs periodically, which has the following significant drawbacks: 1. Data collection methods are crude and lack timeliness. Traditional methods calculate feed intake by manually weighing the difference between the initial and remaining weight of the feed trough at regular intervals. This typically yields the average feed intake for the entire flock on a daily or weekly basis, failing to reflect the instantaneous feeding behavior of individual poultry or capture the temporal distribution patterns of the feeding process. This "static cumulative value" model is severely lagging behind the real-time dynamic data requirements of modern precision farming.

[0004] 2. Difficulty in distinguishing individuals, making individualized monitoring impossible. In group rearing models, manual recording cannot distinguish individual feed intake, resulting in a lack of effective support for applications requiring individual-level data, such as breeding screening and health early warning. Even with electronic feeders, existing systems often lead to data misattribution due to insufficient recognition accuracy or environmental interference.

[0005] 3. Large errors in human operation can easily trigger stress in poultry. Manual weighing requires frequent opening of cages, catching of poultry, and movement of feed troughs, which is not only labor-intensive and inefficient, but also easily causes stress to poultry, affecting their normal feeding behavior and thus reducing the representativeness and reliability of the measurement data.

[0006] 4. Lack of refined dynamic parameters limits behavioral research. Existing technologies can only provide cumulative totals and cannot obtain refined behavioral parameters such as instantaneous feed intake, feeding frequency, feeding duration, and feeding rate of a single pecking, which seriously restricts the development of cutting-edge research fields such as poultry feeding behavior, nutritional metabolism regulation, and feed palatability evaluation.

[0007] 5. Inability to adapt to environmental changes, resulting in insufficient data reliability. The temperature and humidity of aquaculture environments fluctuate greatly, and existing weighing systems generally lack effective temperature drift compensation and nonlinear correction mechanisms, leading to the accumulation of measurement deviations over long-term operation and affecting the long-term stability of the data. At the same time, they lack the ability to intelligently analyze the correlation between environmental parameters (such as temperature and humidity) and feed intake, making it impossible to achieve health early warning and automatic identification of abnormal behavior.

[0008] In summary, current poultry feed intake measurement technology suffers from core bottlenecks such as "low accuracy, limited dimensions, difficulty in individual differentiation, and poor environmental adaptability." There is an urgent need for a high-precision feed intake measurement device and method that can achieve all-time, all-dynamic, individualized, and adaptive measurement to support the transformation and upgrading of modern poultry farming towards data-driven and intelligent decision-making.

[0009] This invention addresses the aforementioned needs by providing a novel poultry feed intake measurement system and method that integrates multi-sensor fusion, three-dimensional spatial positioning, temperature drift compensation, and neural network prediction, effectively filling a gap in existing technology. Summary of the Invention

[0010] To address the aforementioned shortcomings of existing technologies, this invention provides a poultry feed intake measurement device and method, enabling fully dynamic and adaptive measurement of poultry feed intake.

[0011] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: A method for measuring feed intake in poultry is provided, comprising the following steps: S1: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for the weighing sensor. Use the data collected by the weighing sensor to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model, and put them into the data management system. S2: Assign different RFID tags to poultry in poultry cages, and determine whether the poultry has entered the feeding process based on RFID tag identification and temperature field collected by infrared temperature sensors; combine binocular vision recognition and laser ranging technology to determine the effective state of the feeding process and obtain the effective feeding period of poultry. S3: Draw a graph showing the relationship between distance data collected by the laser rangefinder and time during the effective feeding period. Obtain the feeding point based on the maximum value point in the graph, and obtain the weight value collected by the weighing sensor based on the time corresponding to the feeding point to obtain the feed intake data of poultry during the effective feeding period. S4: Calculate the total feed intake and feeding efficiency of poultry during the effective feeding period based on feed intake data, and obtain the time tag for each effective feeding period; S5: Collect farm environmental parameters, poultry weight and breeding age corresponding to the time tag, construct input feature vector, and construct output feature vector using total feed intake and feed intake efficiency; S6: Normalize the input feature vector and the output feature vector to obtain the standardized input feature vector and the standardized output feature vector, which are used as training samples. Use the training samples corresponding to all effective feeding periods during the poultry growth process to construct a training dataset. S7: Construct a BP neural network, including an input layer, hidden layers and an output layer. Input the training samples from the training dataset into the BP neural network, train the BP neural network, correct the network parameters, and output a converged BP neural network. S8: After the data management system calculates the total feed intake and feeding efficiency of poultry during the effective feeding period, it uses a convergent BP neural network to predict the current ideal total feed intake and ideal feeding efficiency of poultry, and evaluates the health status of poultry.

[0012] Further, step S8 includes: S81: The data management system calculates the total feed intake of poultry during the effective feeding period. and feeding efficiency Next, the ambient temperature, humidity, poultry weight, and age of the poultry in the current farm are collected, normalized, and then input into a convergent BP neural network to predict the ideal total feed intake of the poultry. and ideal foraging efficiency ; S82: Calculate total feed intake With ideal total feed intake Difference in feed intake between Feeding efficiency With ideal foraging efficiency Difference in foraging efficiency between Based on the difference in feed intake Difference between feeding efficiency and foraging efficiency Assess the health status of poultry.

[0013] Further, step S1 includes: S11: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for the weighing sensor; ; in, This refers to the weight value collected by the weighing sensor. U This is the voltage signal output by the load cell. This is the zero-point offset coefficient. The coefficients of the linear term, The coefficients of the quadratic nonlinear term are... The coefficients of the cubic nonlinear term, This is the temperature drift error term. t For time, This refers to the real-time temperature of the farm. The reference temperature for calibration. b 0 represents the zero-point temperature drift reference coefficient. b 1 represents the first-order coefficient of zero-point temperature drift. c 0 represents the sensitivity temperature drift reference coefficient. c 1 represents the first-order coefficient of sensitivity temperature drift; S12: After installing the poultry feed intake measuring device on the poultry cages in the farm, different weights of feed are put into different feed troughs and the weighing sensors are calibrated. All weighing sensors continuously collect weight data and output the weight value and the corresponding voltage signal, as well as the temperature data of the farm, as calibration data sets. Several calibration data sets output by all weighing sensors in the farm are obtained. S13: Based on the least squares method, several calibration data sets are used to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model. The fitted third-order nonlinear static calibration model and the temperature drift compensation calibration model are output and put into the data management system.

[0014] Further, step S2 includes: S21: Assign different RFID tags to poultry in poultry cages. The system sets the sensing range of the RFID antenna so that when poultry approach the feed trough, the RFID reader is triggered, and the RFID reader detects the RFID tag. Once the poultry's tag information is identified, the feeding determination process is triggered; The signal strength of the RFID tag is read by the RFID module. To determine whether the RFID reader has been triggered to identify the poultry's tag information; ; in, k Numbering for poultry For time t The trigger state of the RFID reader / writer. The threshold for signal strength, This indicates that the poultry have approached the feed trough, triggering the RFID reader. This indicates that the poultry are not near the feed trough and the RFID reader is not triggered. S22: The infrared temperature sensor collects the temperature field around the feed trough during a continuous period and determines whether poultry temperature data has been collected during the continuous period. If yes, proceed to step S23; otherwise, end the feeding determination process and return to step S21. S23: The binocular vision recognition module acquires images of the feed trough area below and identifies the poultry head area in the image. At the same time, the laser range sensor acquires the vertical distance data of the image area. A three-dimensional coordinate system is constructed with the imaging plane of the image acquired by the binocular vision recognition module as the x0y coordinate plane and the vertical direction as the z-axis. S24: Based on the pixel coordinates of the center pixel of the poultry head region in the image and the vertical distance corresponding to the position of the center pixel, obtain the positioning coordinates of the poultry head region in the three-dimensional coordinate system. ; S25: Based on the coordinates of the feed trough reference point in the three-dimensional coordinate system Calculate the real-time distance between the poultry's head and the feed trough reference point. ; ; S26: Based on real-time distance Construct a model for determining the effective feeding process in poultry; ; in, poultry k In time t The effective state of the foraging process, Indicates poultry k In time t The foraging process is effective. Indicates poultry k In time t The feeding process was invalid. The foraging distance threshold, This represents the change in real-time distance between adjacent time points. The threshold for distance change. The time interval between distance sampling, This represents the real-time distance at the previous time. S27: Based on the effective state of the feeding process of poultry at different times, obtain the effective feeding period of poultry; the feeding process at different times within the effective feeding period is effective.

[0015] Further, step S3 includes: S31: Extract continuous vertical distance data collected by the laser rangefinder during the effective foraging period and plot the relationship between vertical distance and time during the effective foraging period; S32: Take the maximum point in the vertical distance versus time graph as the feeding point when poultry lower their heads, and obtain the time corresponding to each feeding point. , n Numbering the feeding sites; based on the average time poultry spend at the feeding sites. After each feeding, the weighing sensor detects the time... voltage signal Send to the data management system; S33: The data management system outputs voltage signals based on the fitted third-order nonlinear static calibration model and temperature drift compensation calibration model. Corresponding weight value The weight data of the feed trough after each feeding of poultry during the effective feeding period were obtained. ; For the first effective feeding period N Weight of the feed trough after the first feeding; S34: Based on the weight data of the feed trough during the effective feeding period. Calculate the amount of feed consumed by poultry each time. Data on feed intake during each feeding session within the effective feeding period were obtained. ; For the first n -After 1 feeding, For the first N The amount of food consumed during each feeding process.

[0016] Furthermore, step S4 specifically includes: Based on feed intake data Calculate the total feed intake during the effective feeding period. Feeding efficiency ; ; in, m This refers to the numbering of effective feeding periods during the growth of poultry. This is a time label for the feeding process corresponding to the effective feeding period.

[0017] A poultry feed intake measuring device is provided for the above-mentioned poultry feed intake measuring method. It includes an L-shaped support rod installed on a breeding cage, a weighing sensor connected to the lower end of the L-shaped support rod, and a feed trough provided at the lower end of the weighing sensor via an L-shaped bracket. The upper end of the L-shaped support rod is equipped with a binocular vision recognition module, a laser rangefinder, and an infrared temperature sensor. All three components face the feeding point of the feed trough. It also includes a data management system set up in the farm control center and an RFID module set up on the breeding cages. The binocular vision recognition module, laser rangefinder, infrared temperature sensor, weighing sensor and RFID module are all connected to the data management system.

[0018] A farm control center is provided, which includes a display terminal, a processor and a memory. The processor has a data management system built in it. A binocular vision recognition module, a laser rangefinder sensor, an infrared temperature sensor, a weighing sensor and an RFID module are all connected to the data management system. Memory, which is used to store executable instructions; The processor is configured to implement the above-described method for measuring poultry feed intake via executable instructions.

[0019] The beneficial effects of this invention are as follows: This invention enables precise measurement and data analysis of feed intake in poultry farming throughout the entire time period and in real time. It allows for accurate, individualized, real-time data collection of feed intake, and can not only measure static cumulative feed intake but also analyze dynamic parameters across all dimensions, such as single instantaneous feed intake, feed efficiency, feed frequency, and feed duration, in real time. This achieves continuous dynamic monitoring of the poultry feeding process throughout the entire time period, filling a gap in existing technologies. Furthermore, through the integrated neural network model, it can predict real-time reference values ​​for feed intake at various time periods, providing data support for evaluating the health status of poultry.

[0020] In large-scale farms, this technology provides data support for precision nutrition feeding, enabling on-demand feeding, reducing feed consumption, and saving large-scale farms significant feed costs annually. It also replaces manual measurement, reducing labor input and costs. For poultry breeding, it provides high-precision individual feed intake data, significantly improving the efficiency of selecting superior breeds, shortening breeding cycles, and enhancing the market competitiveness of breeding stock. Based on dynamic feed intake monitoring, it enables early warning of poultry diseases, reducing mortality rates and increasing farming profits. Attached Figure Description

[0021] Figure 1 This is a structural diagram of a poultry feed intake measuring device.

[0022] Figure 2 This is a graph showing the relationship between vertical distance and time.

[0023] Figure 3 This is a schematic diagram of a three-dimensional coordinate system.

[0024] The components include: 1. material trough; 2. RFID module; 3. L-shaped bracket; 4. weighing sensor; 5. L-shaped support rod; 6. binocular vision recognition module; 7. laser rangefinder; and 8. infrared temperature sensor. Detailed Implementation

[0025] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0026] A method for measuring feed intake in poultry includes the following steps: S1: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for the weighing sensor 4. Use the data collected by the weighing sensor 4 to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model, and put them into the data management system.

[0027] Step S1 specifically includes the following steps: S11: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for weighing sensor 4; ; in, This is the weight value collected by weighing sensor 4. U This is the voltage signal output by the weighing sensor 4. The zero-point offset coefficient (g) is used. The coefficient of the linear term is (g / mV). The coefficient of the second-order nonlinear term is (g / (mV²)). The coefficient of the cubic nonlinear term is (g / (mV³)). This is the temperature drift error term. t For time, This refers to the real-time temperature of the farm. The reference temperature for calibration is 25°C. b 0 represents the zero-point temperature drift reference factor (g). b 1 represents the first-order coefficient of zero-point temperature drift (g / ℃). c 0 represents the sensitivity temperature drift reference coefficient (g / mV); c 1 represents the first-order coefficient of sensitivity temperature drift. ; S12: After installing the poultry feed intake measuring device on the poultry cages in the farm, different weights of feed are put into different feed troughs 1 and the weighing sensors 4 are calibrated. All weighing sensors 4 continuously collect weight data and output the weight value and the corresponding voltage signal, as well as the temperature data of the farm, as calibration data sets, to obtain several calibration data sets output by all weighing sensors 4 in the farm. S13: Based on the least squares method, several calibration data sets are used to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model. The fitted third-order nonlinear static calibration model and the temperature drift compensation calibration model are output and put into the data management system.

[0028] The output voltage of the load cell 4 is not strictly linear with the actual weight, exhibiting second- and third-order nonlinear distortions, which can lead to excessive errors under both high and low loads. This invention employs third-order polynomial calibration, covering the full-range nonlinear distortion, while also reserving a temperature drift term to provide an interface for subsequent dynamic correction. This significantly reduces the maximum nonlinear error of static calibration, improving the overall measurement accuracy.

[0029] The zero-point drift and sensitivity drift of the weighing sensor 4 both change linearly with temperature. Therefore, the temperature drift error is decomposed into zero-point correlation terms and sensitivity correlation terms. The error of the weighing sensor 4 is calibrated using the actual application environment of the aquaculture farm, and the coefficients are fitted to achieve real-time temperature drift compensation. This solves the measurement drift problem caused by temperature fluctuations (10-35℃) in the aquaculture environment.

[0030] S2: Assign different RFID tags to poultry in the poultry cages. Based on RFID tag identification and temperature field collected by infrared temperature sensor 8, determine whether the feeding process has started. Combine binocular vision recognition and laser ranging technology to determine the effective state of the feeding process and obtain the effective feeding time period of poultry.

[0031] Step S2 specifically includes the following steps: S21: Assign different RFID tags to poultry in poultry cages. The sensing range of the RFID antenna is set so that when poultry approach feed trough 1, the RFID reader is triggered, and the RFID reader detects the RFID tag. Once the poultry's tag information is identified, the feeding determination process is triggered; The signal strength of the RFID tag is read by RFID module 2. To determine whether the RFID reader has been triggered to identify the poultry's tag information; ; in, k Numbering for poultry For time t The trigger state of the RFID reader / writer. The threshold for signal strength, This indicates that the poultry have approached feed trough 1, triggering the RFID reader. This indicates that the poultry are not near feed trough 1 and the RFID reader is not triggered; the effective identification range of the RFID tag is limited by the signal strength threshold to avoid misidentification of poultry at a distance and to achieve unique identification of individual poultry IDs within the poultry cage.

[0032] S22: Infrared temperature sensor 8 collects the temperature field around feed trough 1 within a continuous period of time and determines whether poultry temperature data has been collected within the continuous period of time. If so, proceed to step S23; otherwise, end the feeding determination process and return to step S21. RFID tags on poultry Afterwards, it indicates that the poultry are ready to eat. However, they may also briefly stay and then leave the area near feed trough 1 without eating. Therefore, an infrared temperature field sensor is set up here for a continuous period of time. When the temperature field senses the poultry's temperature data over a longer period of time, it indicates that the poultry have decided to eat in feed trough 1, effectively avoiding misjudgment. In this implementation, the continuous period of time can be set to 5 seconds.

[0033] S23: The binocular vision recognition module 6 acquires images of the area below the feed trough 1 and identifies the poultry head area in the image. At the same time, the laser range sensor 7 acquires the vertical distance data of the image area. A three-dimensional coordinate system is constructed with the imaging plane of the image acquired by the binocular vision recognition module 6 as the x0y coordinate plane and the vertical direction as the z-axis. S24: Based on the pixel coordinates of the center pixel of the poultry head region in the image and the vertical distance corresponding to the position of the center pixel, obtain the positioning coordinates of the poultry head region in the three-dimensional coordinate system. ,like Figure 3 As shown; S25: Based on the coordinates of the feeding reference point of feed trough 1 in the three-dimensional coordinate system Calculate the real-time distance between the poultry's head and the feed trough reference point 1. ; ; By combining binocular vision technology with laser ranging technology, the position of the poultry's head in three-dimensional space is obtained, and then the distance between it and the feeding opening is calculated to determine whether the poultry has entered a valid feeding position.

[0034] S26: Based on real-time distance Construct a model for determining the effective feeding process in poultry; ; in, poultry k In time t The effective state of the foraging process, Indicates poultry k In time t The foraging process is effective. Indicates poultry k In time t The feeding process was invalid. The feeding distance threshold is set to 50mm in this embodiment. This represents the change in real-time distance between adjacent time points. The threshold for distance change is set to 10mm in this embodiment. The time interval between distance sampling, This represents the real-time distance at the previous time. A feeding process can only be considered valid if all three conditions are met simultaneously: "RFID identification is valid", "the head enters the feeding position", and "the head moves within a reasonable range in three-dimensional space (is pecking)".

[0035] S27: Based on the effective state of the feeding process of poultry at different times, obtain the effective feeding period of poultry; the feeding process at different times within the effective feeding period is effective.

[0036] S3: Draw a graph showing the relationship between the distance data collected by the laser rangefinder 7 and time during the effective feeding period. Obtain the feeding point based on the maximum value point in the graph, and obtain the weight value collected by the weighing sensor 4 based on the time corresponding to the feeding point, thereby obtaining the feed intake data of poultry during the effective feeding period.

[0037] Step S3 specifically includes the following steps: S31: Extract continuous vertical distance data collected by laser rangefinder 7 during the effective foraging period, and plot the relationship between vertical distance and time during the effective foraging period, such as... Figure 2 As shown; S32: Take the maximum point in the vertical distance versus time graph as the feeding point when poultry lower their heads, and obtain the time corresponding to each feeding point. , n Numbering the feeding sites; based on the average time poultry spend at the feeding sites. After each feeding, the weighing sensor 4 weighs the food at a certain time. voltage signal Send to the data management system; The poultry's feeding process in feed trough 1 is characterized by frequent head nodding. In the vertical distance versus time graph, each feeding point corresponds to one instance of the poultry lowering its head to feed. The vertical distance represents the distance between the poultry's head and the laser rangefinder 7, and the vertical distance corresponding to each feeding point is the maximum value within that feeding cycle. After the poultry has finished feeding at each feeding point, the interval between periods is determined. Then, weight data is collected to calculate the amount of food consumed per feeding.

[0038] S33: The data management system outputs voltage signals based on the fitted third-order nonlinear static calibration model and temperature drift compensation calibration model. Corresponding weight value We obtained the weight data of feed trough 1 after each feeding of poultry during the effective feeding period. ; For the first effective feeding period N Weight of feed trough 1 after the first feeding; S34: Based on the weight data of feed trough 1 during the effective feeding period. Calculate the amount of feed consumed by poultry each time. Data on feed intake during each feeding session within the effective feeding period were obtained. ; For the first n -Weight of the feed trough after one feeding. For the first N The amount of food consumed during each feeding process.

[0039] S4: Based on feed intake data Calculate the total feed intake during the effective feeding period. Feeding efficiency ; ; in, m This refers to the numbering of effective feeding periods during the growth of poultry. This is a time label for the feeding process corresponding to the effective feeding period.

[0040] The midpoint of the effective feeding period is used as a time marker for the feeding process, which is then used to predict and assess subsequent feed intake in poultry. By statistically analyzing feed intake and efficiency at different effective feeding periods throughout the day, effective data support is provided for assessing the feeding status of poultry. This offers a new technical approach for poultry feeding behavior research, accurately analyzing the feeding behavior patterns of poultry in a single pecking motion, and providing unprecedentedly refined data for poultry appetite regulation and feed palatability evaluation.

[0041] It provides precise data support for poultry nutrition and metabolism research, and can accurately analyze the impact of different nutritional formulas on poultry feeding behavior and feed intake, providing a high-precision measurement method for the development of precision nutrition formulas and the evaluation of feed ingredient efficacy.

[0042] This invention provides core indicators for intelligent early warning of poultry health. Decreased feed intake is the earliest sign of poultry disease. This invention can achieve early warning of diseases and improve the level of biosecurity in poultry farming.

[0043] S5: Collect time stamps Corresponding farm ambient temperature ,humidity Poultry weight breeding age Construct input feature vectors related to the poultry feeding process. ; Utilize total feed intake and feeding efficiency Constructing the output feature vector ; ; ; S6: For the input feature vector and output feature vector Normalization is performed to obtain the standardized input feature vector. Standardized output feature vector , as training samples; construct a training dataset using training samples corresponding to all effective feeding periods during the poultry growth process; S7: Construct a BP neural network, including an input layer, hidden layers and an output layer. Input the training samples from the training dataset into the BP neural network, train the BP neural network, correct the network parameters, and output a converged BP neural network. The linear transformation process from the input layer to the hidden layer is as follows: ; in, These are the weight matrix and bias of the hidden layer, respectively. The linear transformation vector of the hidden layer. The output vector of the hidden layer. For activation functions; The linear transformation process from the hidden layer to the output layer is as follows: ; in, These are the weight matrix and bias of the output layer, respectively. The linear transformation vector of the output layer; Using the Sigmoid activation function to transform the linear vector Mapping to the [0,1] interval yields the predicted feature vector. ; During training, a backpropagation (BP) neural network calculates predicted feature vectors using a loss function. With output feature vector The difference between the values ​​is used to determine the prediction accuracy of the BP neural network. If the prediction accuracy meets the requirements, the BP neural network is output as a converged network. Otherwise, the network parameters of the BP neural network are adjusted and the gradient adjustment method is used to correct the network parameters until the BP neural network converges. The network parameters include the weight matrix and bias of the hidden layer and the output layer.

[0044] S8: After the data management system calculates the total feed intake and feeding efficiency of poultry during the effective feeding period, it uses a convergent BP neural network to predict the current ideal total feed intake and feeding efficiency of poultry, and evaluates the health status of poultry.

[0045] Step S8 specifically includes the following steps: S81: The data management system calculates the total feed intake of poultry during the effective feeding period. and feeding efficiency Next, the ambient temperature, humidity, poultry weight, and age of the poultry in the current farm are collected, normalized, and then input into a convergent BP neural network to predict the ideal total feed intake of the poultry. Feeding efficiency ; S82: Calculate total feed intake With ideal total feed intake Difference in feed intake between Feeding efficiency Feed intake and feeding efficiency Difference in foraging efficiency between Based on the difference in feed intake Difference between feeding efficiency and foraging efficiency Assess the health status of poultry. and If the result is positive, it means that the poultry's current feed intake and feeding efficiency are higher than expected, and its health status is good; otherwise, its health status is poor.

[0046] This invention constructs a BP neural network to predict the feed intake and feed efficiency of poultry at different growth stages and in different breeding environments, thereby assessing the health status of poultry and providing effective data support and feedback for poultry breeding and the regulation of the breeding environment.

[0047] like Figure 1 As shown, a poultry feed intake measuring device is used to perform the above-mentioned poultry feed intake measuring method. It includes an L-shaped support rod 5 installed on a breeding cage. The lower end of the L-shaped support rod 5 is connected to a weighing sensor 4. The lower end of the weighing sensor 4 is provided with a feed trough 1 through an L-shaped bracket 3.

[0048] The upper end of the L-shaped support rod 5 is equipped with a binocular vision recognition module 6, a laser rangefinder 7, and an infrared temperature sensor 8. The binocular vision recognition module 6, the laser rangefinder 7, and the infrared temperature sensor 8 are all facing the feeding point of the feed trough 1.

[0049] It also includes a data management system set up in the farm control center and an RFID module 2 set up on the breeding cages. The binocular vision recognition module 6, laser rangefinder 7, infrared temperature sensor 8, weighing sensor 4 and RFID module 2 are all connected to the data management system.

[0050] A farm control center includes a display terminal, a processor, and a memory. The processor has a data management system built in it. The binocular vision recognition module 6, the laser rangefinder sensor 7, the infrared temperature sensor 8, the weighing sensor 4, and the RFID module 2 are all connected to the data management system. Memory, which is used to store executable instructions; The processor is configured to implement the above-described method for measuring poultry feed intake via executable instructions.

Claims

1. A method for determining feed intake in poultry, characterized in that, Includes the following steps: S1: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for the weighing sensor. Use the data collected by the weighing sensor to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model, and put them into the data management system. S2: Assign different RFID tags to poultry in poultry cages, and determine whether the poultry has entered the feeding process based on RFID tag identification and temperature field collected by infrared temperature sensors; combine binocular vision recognition and laser ranging technology to determine the effective state of the feeding process and obtain the effective feeding period of poultry. S3: Draw a graph showing the relationship between distance data collected by the laser rangefinder and time during the effective feeding period. Obtain the feeding point based on the maximum value point in the graph, and obtain the weight value collected by the weighing sensor based on the time corresponding to the feeding point to obtain the feed intake data of poultry during the effective feeding period. S4: Calculate the total feed intake and feeding efficiency of poultry during the effective feeding period based on feed intake data, and obtain the time tag for each effective feeding period; S5: Collect farm environmental parameters, poultry weight and breeding age corresponding to the time tag, construct input feature vector, and construct output feature vector using total feed intake and feed intake efficiency; S6: Normalize the input feature vector and the output feature vector to obtain the standardized input feature vector and the standardized output feature vector, which are used as training samples. Use the training samples corresponding to all effective feeding periods during the poultry growth process to construct a training dataset. S7: Construct a BP neural network, including an input layer, hidden layers and an output layer. Input the training samples from the training dataset into the BP neural network, train the BP neural network, correct the network parameters, and output a converged BP neural network. S8: After the data management system calculates the total feed intake and feeding efficiency of poultry during the effective feeding period, it uses a convergent BP neural network to predict the current ideal total feed intake and ideal feeding efficiency of poultry, and evaluates the health status of poultry.

2. The method for determining poultry feed intake according to claim 1, characterized in that, Step S8 includes: S81: The data management system calculates the total feed intake of poultry during the effective feeding period. and feeding efficiency Next, the ambient temperature, humidity, poultry weight, and age of the poultry in the current farm are collected, normalized, and then input into a convergent BP neural network to predict the ideal total feed intake of the poultry. and ideal foraging efficiency ; S82: Calculate total feed intake With ideal total feed intake Difference in feed intake between Feeding efficiency With ideal foraging efficiency Difference in foraging efficiency between Based on the difference in feed intake Difference in foraging efficiency Evaluate the health status of poultry.

3. The method for determining poultry feed intake according to claim 1, characterized in that, Step S1 includes: S11: Construct a third-order nonlinear static calibration model and a temperature drift compensation calibration model for the weighing sensor; ； in, This refers to the weight value collected by the weighing sensor. U This is the voltage signal output by the load cell. This is the zero-point offset coefficient. The coefficients of the linear term, The coefficients of the quadratic nonlinear term are... The coefficients of the cubic nonlinear term, This is the temperature drift error term. t For time, This refers to the real-time temperature of the farm. The reference temperature for calibration. b 0 represents the zero-point temperature drift reference coefficient. b 1 represents the first-order coefficient of zero-point temperature drift. c 0 represents the sensitivity temperature drift reference coefficient. c 1 represents the first-order coefficient of sensitivity temperature drift; S12: After installing the poultry feed intake measuring device on the poultry cages in the farm, different weights of feed are put into different feed troughs and the weighing sensors are calibrated. All weighing sensors continuously collect weight data and output the weight value and the corresponding voltage signal, as well as the temperature data of the farm, as calibration data sets. Several calibration data sets output by all weighing sensors in the farm are obtained. S13: Based on the least squares method, several calibration data sets are used to fit the third-order nonlinear static calibration model and the temperature drift compensation calibration model. The fitted third-order nonlinear static calibration model and the temperature drift compensation calibration model are output and put into the data management system.

4. The method for determining poultry feed intake according to claim 3, characterized in that, Step S2 includes: S21: Assign different RFID tags to poultry in poultry cages. The system sets the sensing range of the RFID antenna so that when poultry approach the feed trough, the RFID reader is triggered, and the RFID reader detects the RFID tag. Once the poultry's tag information is identified, the feeding determination process is triggered; The signal strength of the RFID tag is read by the RFID module. To determine whether the RFID reader has been triggered to identify the poultry's tag information; ； in, k Numbering for poultry For time t The trigger state of the RFID reader / writer. The threshold for signal strength, This indicates that the poultry have approached the feed trough, triggering the RFID reader. This indicates that the poultry are not near the feed trough and the RFID reader is not triggered. S22: The infrared temperature sensor collects the temperature field around the feed trough during a continuous period and determines whether poultry temperature data has been collected during the continuous period. If yes, proceed to step S23; otherwise, end the feeding determination process and return to step S21. S23: The binocular vision recognition module acquires images of the feed trough area below and identifies the poultry head area in the image. At the same time, the laser range sensor acquires the vertical distance data of the image area. A three-dimensional coordinate system is constructed with the imaging plane of the image acquired by the binocular vision recognition module as the x0y coordinate plane and the vertical direction as the z-axis. S24: Based on the pixel coordinates of the center pixel of the poultry head region in the image and the vertical distance corresponding to the position of the center pixel, obtain the positioning coordinates of the poultry head region in the three-dimensional coordinate system. ; S25: Based on the coordinates of the feed trough reference point in the three-dimensional coordinate system Calculate the real-time distance between the poultry's head and the feed trough reference point. ; ； S26: Based on real-time distance Construct a model for determining the effective feeding process in poultry; ； in, poultry k In time t The effective state of the foraging process, Indicates poultry k In time t The foraging process is effective. Indicates poultry k In time t The feeding process was invalid. The foraging distance threshold, This represents the change in real-time distance between adjacent time points. The threshold for distance change. The time interval between distance sampling, This represents the real-time distance at the previous time point; S27: Based on the effective state of the feeding process of poultry at different times, obtain the effective feeding period of poultry; the feeding process at different times within the effective feeding period is effective.

5. The method for determining poultry feed intake according to claim 4, characterized in that, Step S3 includes: S31: Extract continuous vertical distance data collected by the laser rangefinder during the effective foraging period and plot the relationship between vertical distance and time during the effective foraging period; S32: Take the maximum point in the vertical distance versus time graph as the feeding point of the poultry, and obtain the time corresponding to each feeding point. , n Numbering the feeding sites; based on the average time poultry spend at the feeding sites. After each feeding, the weighing sensor detects the time... voltage signal Send to the data management system; S33: The data management system outputs voltage signals based on the fitted third-order nonlinear static calibration model and temperature drift compensation calibration model. Corresponding weight value The weight data of the feed trough after each feeding of poultry during the effective feeding period were obtained. ; For the first effective feeding period N Weight of the feed trough after the first feeding; S34: Based on the weight data of the feed trough during the effective feeding period. Calculate the amount of feed consumed by poultry each time. Data on feed intake during each feeding session within the effective feeding period were obtained. ; For the first n -After 1 feeding, For the first N The amount of food consumed during each feeding process.

6. The method for determining poultry feed intake according to claim 5, characterized in that, Step S4 specifically involves: Based on feed intake data Calculate the total feed intake during the effective feeding period. Feeding efficiency ; ； in, m This refers to the numbering of effective feeding periods during the growth of poultry. This is a time label for the feeding process corresponding to the effective feeding period.

7. A poultry feed intake measuring device, used to perform the poultry feed intake measuring method according to any one of claims 1-6, characterized in that, It includes an L-shaped support rod installed on the breeding cage, with a weighing sensor connected to the lower end of the L-shaped support rod, and a feed trough set at the lower end of the weighing sensor via an L-shaped bracket; The upper end of the L-shaped support rod is equipped with a binocular vision recognition module, a laser rangefinder, and an infrared temperature sensor. All three components face the feeding point of the feed trough. It also includes a data management system set up in the farm control center and an RFID module set up on the breeding cages. The binocular vision recognition module, laser rangefinder, infrared temperature sensor, weighing sensor and RFID module are all connected to the data management system.

8. A farm control center, characterized in that, It includes a display terminal, a processor, and a memory. The processor has a data management system built in it. The binocular vision recognition module, laser rangefinder, infrared temperature sensor, weighing sensor, and RFID module are all communicatively connected to the data management system. The memory is used to store executable instructions; The processor is configured to implement the poultry feed intake measurement method according to any one of claims 1-6 via the executable instructions.

Images