Self-adaptive precise feeding method for feed conveying line

Abstract

The invention discloses a feed delivery line self-adaptive accurate feeding method, and belongs to the technical field of feeding, and the method comprises the steps: collecting chickens and chicken flocks to obtain multi-modal data and chicken behavior data; constructing a social network graph of the chicken flocks according to the chicken behavior data to obtain individual indexes of each chicken, obtaining fusion data based on the multi-modal data, the chicken behavior data and the individual indexes, inputting the fusion data into a multi-modal time sequence prediction model to obtain growth trend prediction data of each chicken, and predicting the growth trend of each chicken according to the growth trend prediction data. A composite feeding strategy is obtained based on the growth trend prediction data and the feed consumption, and the feed conveying line unit is regulated and controlled to conduct feeding according to the composite feeding strategy. According to the method, the dynamic nutritional requirements of chicken flocks in different growth stages, different environment temperatures and different health states can be met, the individual difference between different chickens is met by predicting the growth of each chicken, the laying rate of the chickens is increased, and the feeding amount of feed can be adaptively adjusted according to a composite feeding strategy.

Description

Technical Field

[0001] This invention belongs to the field of feeding technology, specifically relating to an adaptive precision feeding method for a feed line. Background Technology

[0002] Currently, many chicken houses in my country still use traditional manual feeding methods. Feeding refers to the operation of providing feed to animals according to feeding standards, which has certain standardized requirements, such as timed feeding, quantitative feeding, quality feeding, and frequent small feedings. It also requires a reasonable combination of concentrates and roughage. Quantitative feeding involves feeding a relatively stable amount of feed within a certain period to meet the nutritional needs of each chicken, avoiding unnecessary losses and waste. Quality feeding requires that the feed be fresh and clean, ensuring that the quality and nutritional cost meet requirements, and ensuring that rotten or moldy feed is not fed. Based on the digestive characteristics of the flock and the types of feed, a variety of raw materials should be selected to enhance nutritional complementarity and prevent nutritional deficiencies.

[0003] However, the existing chicken house feeding methods still have problems such as serious feed waste, uneven growth of chickens and inability to make dynamic adjustments. Specifically, these include: (1) serious feed waste: the existing artificial feeding method uses fixed time and fixed amount to feed, ignoring the dynamic nutritional needs of chickens at different growth stages, different environmental temperatures and different health states, resulting in serious feed waste or insufficient nutrient supply; (2) uneven growth of chickens: all chickens receive the same amount of feed, which cannot meet the needs of individual differences, resulting in uneven growth of chickens and difficulty for weak chickens to eat, leading to an unsatisfactory overall feed-to-meat ratio or feed-to-egg ratio; (3) inability to make dynamic adjustments: the traditional chicken house feeding method cannot accurately know the amount of feed, feeding behavior and health status of chickens after feeding, and cannot make dynamic adjustments according to the existing situation.

[0004] Therefore, how to provide an effective technical solution to address the problems of serious feed waste, uneven growth of chicken flocks, and inability to dynamically adjust existing technologies has become an urgent problem to be solved in existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive precision feeding method for a feed line to solve the above-mentioned problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides an adaptive precision feeding method for a feeder line, comprising: The system acquires multimodal data collected by multimodal sensors deployed in the chicken house and video sequences synchronously collected by wide-angle cameras deployed in the chicken house. The multimodal data includes weight data, body temperature data, and environmental data. Based on video sequences, chicken behavior data is obtained. Using graph neural networks, a social network graph of the chicken flock is constructed based on the chicken behavior data. Individual indicators of each chicken are obtained based on the social network graph. Multimodal data, chicken behavior data, and individual indicators are fused to obtain fused data. The fused data is then input into a multimodal time series prediction model to obtain growth trend prediction data for each chicken within a preset future time period. The growth trend prediction data includes health trend, growth curve deviation risk, and potential disease risk. The feed consumption is monitored in real time. Based on the multi-objective optimization algorithm, the growth trend prediction data and feed consumption of each chicken within a preset time period are used to generate a strategy to obtain a compound feeding strategy. The compound feeding strategy includes the target feeding amount, feeding time window and feed formula ratio. The compound feeding strategy is sent to the control center so that the control center can generate control instructions based on the compound feeding strategy and regulate the feeding line unit to feed the corresponding feed trough based on the control instructions.

[0007] In one possible design, after controlling the feeding line unit to feed the corresponding trough based on control commands, the following is also included: Obtain the feed value detected by the pressure sensor installed at the bottom of the feed trough; The compound feeding strategy is validated based on feed values ​​to obtain validation results. If the validation results are correct, the execution results after feeding are obtained, which are the feed consumption after feeding. The multi-objective optimization algorithm is updated based on the execution results.

[0008] In one possible design, chicken behavior data is obtained through video sequence recognition. Based on this data, a graph neural network is used to construct a social network graph of the chicken flock, including: Each chicken in the video sequence is identified, and its movement trajectory is tracked within a preset time period. The interaction events between each chicken are statistically analyzed based on its movement trajectory to obtain the interaction event statistics. The interaction events include approaching, following, chasing, and pecking. The interaction event statistics are used as chicken behavior data. Based on the statistical results of interactive events, weights are assigned to the connection edges between individual chickens. Using a graph neural network, a weighted directed graph is constructed based on the connection edges with different weights and individual chickens. The centrality index of each node in the weighted directed graph is calculated to obtain the centrality index of each node, and a social network graph of the chicken flock is constructed based on the centrality index of each node.

[0009] In one possible design, the multimodal time-series prediction model is constructed based on the Transformer model; the step of inputting the fused data into the multimodal time-series prediction model to obtain the growth trend prediction data of each chicken within a preset future time period includes: Get the preset time period in the future; The encoder is used to perform deep feature extraction on the fused data to obtain fused features, which include weight features, body temperature features, environmental features, behavioral features and index features. The decoder uses a fusion feature and attention mechanism to predict the growth trend of each chicken within a preset future time period.

[0010] In one possible design, before fusing multimodal data, chicken behavior data, and individual metrics, the following is also included: Data cleaning and outlier handling are performed on the multimodal data to obtain the processed multimodal data; The processed multimodal data is denoised using an adaptive filtering algorithm to obtain denoised multimodal data. The denoised multimodal data is normalized.

[0011] In one possible design, after acquiring the multimodal data collected by the multimodal sensors deployed in the chicken coop and the video sequence synchronously acquired by the wide-angle camera deployed in the chicken coop, the following is also included: Frame processing is performed on the video sequence to obtain a sequence of chicken frame images; Based on a pre-built video analysis model, the state of the chicken flock is analyzed in the chicken frame image sequence to obtain the flock state, which includes the overall distribution of the flock, the activity level of the chickens, and the feather state of the chickens. The health status of the flock is determined based on the overall distribution of the flock, the activity level of the chickens, and the condition of their feathers. If the health status of the flock is lower than the preset health level, an early warning plan is generated based on the pre-built database. The pre-built digital twin model is updated in real time based on the health status of the flock.

[0012] In one possible design, after updating the pre-built digital twin model in real time based on the flock's health status, it also includes: Based on the updated digital twin model, an adaptive feeding strategy generation algorithm is used to generate an updated feeding strategy. The updated feeding strategy parameters are sent to the control center so that the control center can update the compound feeding strategy based on the updated feeding strategy parameters, obtain the feeding update control command, and regulate the feeding line unit to feed the corresponding feed trough based on the feeding update control command.

[0013] Secondly, the present invention provides an adaptive precision feeding device for a feed line, comprising: The information acquisition unit is used to acquire multimodal data collected by multimodal sensors deployed in the chicken house and video sequences synchronously acquired by wide-angle cameras deployed in the chicken house. The multimodal data includes weight data, body temperature data and environmental data. The graph construction unit is used to identify chicken behavior data based on video sequences. Then, a graph neural network is used to construct a social network graph of the chicken flock based on the chicken behavior data, and individual indicators of each chicken are obtained based on the social network graph. The trend prediction unit is used to fuse multimodal data, chicken behavior data and individual indicators to obtain fused data. The fused data is then input into the multimodal time series prediction model to obtain the growth trend prediction data of each chicken within a preset time period. The growth trend prediction data includes health trend, growth curve deviation risk and potential disease risk. The strategy generation unit is used to obtain the real-time monitored feed consumption, and generate a strategy based on the growth trend prediction data and feed consumption of each chicken within a preset time period using a multi-objective optimization algorithm, thereby obtaining a compound feeding strategy. The compound feeding strategy includes the target feeding amount, feeding time window and feed formula ratio. The feeding control unit is used to send the compound feeding strategy to the control center, so that the control center can generate control instructions based on the compound feeding strategy and regulate the feeding line unit to feed the corresponding feed trough based on the control instructions.

[0014] Thirdly, the present invention provides a computer device comprising a memory, a processor, and a transceiver connected in sequence and communication, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the adaptive precision feeding method for the feed line as described in the first aspect.

[0015] Fourthly, the present invention provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the adaptive precision feeding method for a feed line as described in the first aspect.

[0016] The beneficial effects of this invention are as follows: This invention discloses an adaptive precision feeding method for feed lines, including acquiring multimodal data collected by multimodal sensors deployed in the chicken house and video sequences synchronously collected by wide-angle cameras deployed in the chicken house. The multimodal data includes weight data, body temperature data, and environmental data, and is identified based on the video sequences to obtain chicken behavior data; using a graph neural network, a social network graph of the chicken flock is constructed based on the chicken behavior data, and individual indicators of each chicken are obtained based on the social network graph; the multimodal data, chicken behavior data, and individual indicators are fused to obtain fused data; the fused data is input into a multimodal time series prediction model to obtain growth trend prediction data for each chicken within a preset future time; a strategy is generated based on the growth trend prediction data and feed consumption of each chicken within the preset future time using a multi-objective optimization algorithm to obtain a compound feeding strategy; the compound feeding strategy is sent to the control center so that the control center can generate control instructions based on the compound feeding strategy and regulate the feed line unit to feed the corresponding feed trough based on the control instructions. This invention predicts the growth trend of each chicken and generates a compound feeding strategy based on feed consumption. This strategy can meet the dynamic nutritional needs of chickens at different growth stages, in different environmental temperatures, and under different health conditions, avoiding serious feed waste or insufficient nutrient supply. By predicting the growth of each chicken, it addresses individual differences among chickens and avoids uneven growth that could lead to an unsatisfactory overall feed conversion ratio or feed-to-egg ratio, thereby improving the egg production rate. The compound feeding strategy can adaptively adjust the amount of feed, making it easy to apply and promote. Attached Figure Description

[0017] Figure 1 A flowchart illustrating an adaptive precision feeding method for a feed line provided in an embodiment of the present invention; Figure 2 This is a structural diagram of an adaptive precision feeding device for a feed line provided in an embodiment of the present invention; Figure 3 This is a structural diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is 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. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0019] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0020] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0021] Example: like Figure 1 As shown, the first aspect of this embodiment provides an adaptive precision feeding method for a feed line, which can be executed, but is not limited to, by a computer device or virtual machine with certain computing resources, such as a personal computer or smartphone, or by a virtual machine; the adaptive precision feeding method for a feed line includes, but is not limited to, the following steps: S1. Acquire multimodal data collected by multimodal sensors deployed in the chicken house and video sequences synchronously collected by wide-angle cameras deployed in the chicken house. The multimodal data includes weight data, body temperature data, and environmental data. It should be noted that multimodal sensors include, but are not limited to, weight detection sensors and temperature sensors. These can be deployed along the chickens' usual routes in the chicken coop to reduce measurement costs. For example, installing weight detection sensors and temperature sensors on each chicken's drinking or foraging route eliminates the need for manual weighing and temperature measurement of each chicken, thus reducing labor costs. Simultaneously, a unique identification identifier for each chicken is obtained through an RFID reader. The multimodal data is then associated with the unique identification identifier and uploaded to a central processing platform to establish and continuously update a health record for each chicken. In one possible implementation, after acquiring the multimodal data collected by the multimodal sensors deployed in the chicken coop and the video sequence synchronously acquired by the wide-angle camera deployed in the chicken coop, the method further includes: The video sequence is processed frame by frame to obtain a chicken frame image sequence. Based on a pre-built video analysis model, the chicken flock status is analyzed to obtain the flock status, which includes the overall distribution of the flock, the activity level of the chickens, and the feather condition of the chickens. The health level of the flock is judged based on the overall distribution of the flock, the activity level of the chickens, and the feather condition of the chickens. If the health level of the flock is lower than the preset health level, an early warning scheme is generated based on the pre-built database. The pre-built digital twin model is updated in real time based on the health level of the flock.

[0022] It should be noted that the video analysis model is built on convolutional neural networks or recurrent neural networks. After training the video analysis model with a large amount of labeled data, it can analyze the status of chicken flocks and automatically classify the feather status of chickens. By analyzing the overall distribution of the flock, the activity level of the chickens, and the feather status of the chickens, it can determine the health level of the flock and obtain a preset health level table. The model compares the determined health level of the flock with the preset health level table. If the health level of the flock is lower than the health level in the preset health level table, it generates a corresponding early warning plan based on a pre-built database, such as increasing the ambient temperature of the chicken house. The digital twin model is a virtual mirror built on the physical system of the chicken house. By accessing multimodal sensors and video streams in real time, it dynamically maps the flock status, environmental parameters, and equipment operation. This model can reflect the current situation and simulate the growth, health, and feeding behavior of the flock under different feeding strategies based on historical and real-time data. This allows for strategy prediction and optimization in a virtual environment, achieving intelligent closed-loop management and improving feeding accuracy.

[0023] Furthermore, after updating the pre-built digital twin model in real time based on the flock's health status, it also includes: Based on the updated digital twin model, an adaptive feeding strategy generation algorithm is used to generate updated feeding strategy parameters. These updated parameters are then sent to the control center, which updates the composite feeding strategy based on the updated parameters, obtains feeding update control commands, and regulates the feed line unit to feed the corresponding feed troughs based on these commands.

[0024] Among them, the adaptive feeding strategy generation algorithm is based on deep reinforcement learning algorithm. Its principle is to simulate and optimize the strategy in a virtual digital twin environment by collecting multimodal data of chicken flocks in real time and predicting growth trends. The strategy effect is evaluated by using reward functions (such as growth efficiency, uniformity or feed cost), and the algorithm continuously iterates through reinforcement learning to generate dynamically adapted feeding parameters (feeding amount, formula and / or time).

[0025] S2. Based on video sequences, identify chicken behavior data, use graph neural networks to construct a social network graph of the chicken flock based on the chicken behavior data, and obtain individual indicators for each chicken based on the social network graph; Specifically, in step S2, chicken behavior data is obtained based on video sequences. Then, a graph neural network is used to construct a social network graph of the chicken flock based on this behavior data, including: S21. Identify each chicken in the video sequence and track the movement trajectory of each chicken within a preset time period; S22. Based on the movement trajectory, the interaction events between each chicken are statistically analyzed to obtain the interaction event statistics results. The interaction events include approaching, following, chasing and pecking. The interaction event statistics results are used as chicken behavior data. S23. Assign weights to the connection edges between individual chickens based on the statistical results of the interaction events, and use a graph neural network to construct a weighted directed graph based on the connection edges with different weights and the individual chickens. S24. Calculate the centrality index of each node in the weighted directed graph to obtain the centrality index of each node, and construct the social network graph of the chicken flock based on the centrality index of each node.

[0026] It should be noted that the YOLOv8 model is used to identify each chicken in the video sequence, and the DeepSort algorithm is used to track the movement trajectory of each chicken within a preset time period. The principle of the YOLOv8 model is to treat the object detection problem as a regression problem. The input image is divided into a grid, and each grid predicts multiple bounding boxes. Finally, the results are filtered by non-maximum suppression. The YOLOv8 model is based on the single-stage detection idea. It predicts the bounding boxes and class probabilities of objects in the image in one go through an end-to-end architecture. It achieves a balance between speed and accuracy through efficient single forward propagation, modular network design, and AnchorFree mechanism. The DeepSort algorithm is a multi-object tracking algorithm. It introduces deep appearance features and cascaded matching mechanism on the basis of the Sort algorithm, which can significantly improve the tracking accuracy in target occlusion or similar scenes. The core principle is to fuse motion prediction and appearance similarity for data association.

[0027] S3. Perform data fusion on multimodal data, chicken behavior data and individual indicators to obtain fused data. Input the fused data into the multimodal time series prediction model to obtain the growth trend prediction data of each chicken in the future within a preset time. The growth trend prediction data includes health trend, growth curve deviation risk and potential disease risk. It should be noted that the multimodal time series prediction model is built based on the Transformer model. Before fusing multimodal data, chicken behavior data, and individual indicators, the multimodal data, chicken behavior data, and individual indicators are standardized and aligned to facilitate data fusion of different data types. Data fusion of multimodal data, chicken behavior data, and individual indicators can be performed using machine learning-based data fusion methods, which are existing technologies and will not be elaborated on here.

[0028] Specifically, in step S3, the fused data is input into the multimodal time-series prediction model to obtain the growth trend prediction data for each chicken within a preset future time period, including: S31. Obtain the preset time period in the future; S32. Use an encoder to perform deep feature extraction on the fused data to obtain fused features, which include weight features, body temperature features, environmental features, behavioral features, and indicator features; S33. Using a decoder, based on fusion features and an attention mechanism, predictions are made for the growth trend of each chicken within a future preset time period to obtain the predicted data.

[0029] In a preferred embodiment, before fusing the multimodal data, chicken behavior data, and individual indicators, the method further includes: Data cleaning and outlier handling are performed on the multimodal data to obtain processed multimodal data; the processed multimodal data is then denoised using an adaptive filtering algorithm to obtain denoised multimodal data; and the denoised multimodal data is then normalized.

[0030] By cleaning, denoising, and normalizing multimodal data, the data types are unified, which facilitates subsequent data fusion and model processing.

[0031] S4. Obtain the real-time monitored feed consumption, and generate a strategy based on the growth trend prediction data and feed consumption of each chicken within a preset time period using a multi-objective optimization algorithm to obtain a compound feeding strategy. The compound feeding strategy includes the target feeding amount, feeding time window and feed formula ratio. In a preferred embodiment, a compound feeding strategy is generated based on the predicted growth trend data and feed consumption of each chicken within a preset future time period using a multi-objective optimization algorithm, including: S41. Define an objective function based on the predicted growth trend data and feed consumption within a preset future time period. The objective function includes a growth performance function, a population evenness function, and a feed cost function. S42. Obtain preset constraints, including feeding amount threshold, minimum nutritional index and maximum feed inventory; S43. An initial strategy population is randomly generated based on a multi-objective optimization algorithm and preset constraints, wherein the initial strategy population includes multiple initial composite feeding strategies; S44. Input the initial strategy population into the multimodal time series prediction model to obtain the prediction results, which include the predicted growth trend data of each chicken in the future when using a certain initial compound feeding strategy in the initial strategy population. S45. Based on the prediction results, obtain multiple target scores for the corresponding initial policy population in the objective function; S46. A multi-objective optimization algorithm is used to sort the scores of multiple objectives to obtain the Pareto front. The Pareto front is the set of all unmanageable solutions.

[0032] S47. Perform evolutionary operations on the Pareto front to obtain a compound feeding strategy. The evolutionary operations include, but are not limited to, selection, crossover, and mutation.

[0033] In practice, evolutionary operations are performed on the Pareto front, including: selecting an initial compound feeding strategy from the Pareto front and using the selected initial compound feeding strategy as the parent strategy; combining the decision variables in the parent strategy to obtain the child strategy; merging the parent strategy and the child strategy to obtain the merged strategy population; repeating iterative steps S43-S47 until a preset number of iterations is reached or the Pareto front no longer improves significantly, to obtain the Pareto optimal strategy set, which is then used as the compound feeding strategy.

[0034] It should be noted that, specifically, combining the decision variables in the parent strategy to obtain the sub-strategy can be exemplified by combining the feeding amount of strategy A with the feed formula of strategy B.

[0035] S5. The control center generates control commands based on the compound feeding strategy, and adjusts the feed line unit to feed the corresponding feed trough based on the control commands.

[0036] In one possible implementation, after controlling the feed line unit to feed the corresponding trough based on control commands, the method further includes: The feed value detected by the pressure sensor installed at the bottom of the feed trough is obtained; the compound feeding strategy is verified based on the feed value, and the verification result is obtained. If the verification result is correct, the execution result after feeding is obtained, which is the feed consumption after feeding. The multi-objective optimization algorithm is updated based on the execution result.

[0037] It should be noted that the verification here is mainly to prove whether the actual feeding amount of the feed line is consistent with the target feeding amount set in the compound feeding strategy.

[0038] Based on the above disclosure, this embodiment provides an adaptive precision feeding device for a feed line. By collecting individual and group information of chickens, chicken behavior data is obtained. A social network graph of the chicken flock is constructed based on the chicken behavior data, and individual indicators of each chicken are obtained. This predicts the growth trend of each chicken within a preset time period. A compound feeding strategy is generated based on feed consumption and growth trend to meet the dynamic and personalized needs of chickens at different growth stages, different environmental temperatures, and different health states. An early warning mechanism is introduced to analyze the health status of each chicken and generate a corresponding early warning plan based on the health status of each chicken, thereby adjusting the corresponding settings.

[0039] like Figure 2 As shown, the second aspect of this embodiment provides an adaptive precision feeding device for a feed line, comprising: The information acquisition unit is used to acquire multimodal data collected by multimodal sensors deployed in the chicken house and video sequences synchronously acquired by wide-angle cameras deployed in the chicken house. The multimodal data includes weight data, body temperature data and environmental data. The graph construction unit is used to identify chicken behavior data based on video sequences. Then, a graph neural network is used to construct a social network graph of the chicken flock based on the chicken behavior data, and individual indicators of each chicken are obtained based on the social network graph. The trend prediction unit is used to fuse multimodal data, chicken behavior data and individual indicators to obtain fused data. The fused data is then input into the multimodal time series prediction model to obtain the growth trend prediction data of each chicken within a preset time period. The growth trend prediction data includes health trend, growth curve deviation risk and potential disease risk. The strategy generation unit is used to obtain the real-time monitored feed consumption, and generate a strategy based on the growth trend prediction data and feed consumption of each chicken within a preset time period using a multi-objective optimization algorithm, thereby obtaining a compound feeding strategy. The compound feeding strategy includes the target feeding amount, feeding time window and feed formula ratio. The feeding control unit sends the compound feeding strategy to the control center, so that the control center can generate control instructions based on the compound feeding strategy and regulate the feeding line unit to feed the corresponding feed trough based on the control instructions.

[0040] The working process, working details and technical effects of the adaptive precision feeding device for a feed line provided in the second aspect of this embodiment can be found in the adaptive precision feeding method for a feed line described in the first aspect, and will not be repeated here.

[0041] like Figure 3 As shown, the third aspect of this embodiment provides a computer device, including a memory, a processor, and a transceiver connected in sequence for communication. The memory stores a computer program, the transceiver sends and receives messages, and the processor reads the computer program to execute the adaptive precision feeding method for feed lines as described in the first aspect. Specifically, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. Furthermore, the computer device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0042] The working process, working details and technical effects of the aforementioned computer equipment provided in the third aspect of this embodiment can be found in the adaptive precision feeding method for the feed line described in the first aspect, and will not be repeated here.

[0043] The fourth aspect of this embodiment provides a computer program product, including a computer program or instructions, which, when executed by a computer, are used to implement the adaptive precision feeding method for the feed line as described in the first aspect.

[0044] The working process, working details and technical effects of the aforementioned computer program product provided in this embodiment can be found in the adaptive precision feeding method for feed lines as described in the first aspect, and will not be repeated here.

[0045] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for adaptive precision feeding of a feed line, characterized in that, The method comprises the following steps: acquiring multi-modal data collected by multi-modal sensors deployed in a chicken house and video sequences synchronously collected by wide-angle cameras deployed in the chicken house, wherein the multi-modal data comprises body weight data, body temperature data and environmental data; recognizing the video sequences to obtain chicken behavior data, constructing a social network graph of the chicken group based on the chicken behavior data using a graph neural network, and obtaining individual indexes of each chicken based on the social network graph; performing data fusion on the multi-modal data, the chicken behavior data and the individual indexes to obtain fused data, inputting the fused data into a multi-modal time series prediction model to obtain growth trend prediction data of each chicken within a preset future time, wherein the growth trend prediction data comprises health trend, growth curve deviation risk and potential disease risk; acquiring real-time monitored feed consumption, generating a compound feeding strategy based on a multi-objective optimization algorithm for the growth trend prediction data of each chicken within the preset future time and the feed consumption, wherein the compound feeding strategy comprises a target feeding amount, a feeding time window and a feed formula ratio; sending the compound feeding strategy to a control center so that the control center generates a control instruction based on the compound feeding strategy and regulates a feed line unit to feed corresponding troughs based on the control instruction.

2. The method according to claim 1, wherein, After regulating the feed line unit to feed the corresponding troughs based on the control instruction, the method further comprises the following steps: acquiring a feed value detected by a pressure sensor installed at the bottom of the trough; verifying the compound feeding strategy based on the feed value to obtain a verification result, if the verification result is correct, acquiring an execution result after feeding, wherein the execution result is the feed consumption after feeding, and updating the multi-objective optimization algorithm based on the execution result.

3. The method according to claim 1, wherein, The method of recognizing the video sequences to obtain chicken behavior data and constructing a social network graph of the chicken group based on the chicken behavior data using a graph neural network comprises the following steps: recognizing each chicken in the video sequences and tracking the movement trajectory of each chicken within a preset time period; counting the interaction events between each chicken based on the movement trajectory to obtain an interaction event statistical result, wherein the interaction events comprise approaching, following, driving and pecking, and the interaction event statistical result is used as the chicken behavior data; assigning weights to the connection edges between chicken individuals according to the interaction event statistical result, constructing a weighted directed graph based on the connection edges with different weights and the chicken individuals using a graph neural network; calculating the centrality index of each node of the weighted directed graph to obtain the centrality index of each node, and constructing a social network graph of the chicken group based on the centrality index of each node.

4. The method according to claim 1, wherein, The multi-modal time series prediction model is constructed based on a Transformer model; and the method of inputting the fused data into the multi-modal time series prediction model to obtain the growth trend prediction data of each chicken within the preset future time comprises the following steps: acquiring a preset future time period; extracting deep features of the fused data using an encoder to obtain fused features, wherein the fused features comprise body weight features, body temperature features, environmental features, behavior features and index features; The decoder is used to predict the growth trend prediction data of each chicken in the future preset time period based on the fusion features and the attention mechanism.

5. The method of claim 1, wherein, Before data fusion is performed on the multi-modal data, the chicken behavior data, and the individual indicators, the following operations are further included: Data cleaning and outlier processing operations are performed on the multi-modal data to obtain processed multi-modal data. The processed multi-modal data is denoised based on an adaptive filtering algorithm to obtain denoised multi-modal data. The denoised multi-modal data is normalized.

6. The method of claim 1, wherein, After the multi-modal data collected by the multi-modal sensors deployed in the chicken house and the video sequences synchronously collected by the wide-angle cameras deployed in the chicken house are obtained, the following operations are further included: Frame processing is performed on the video sequences to obtain a sequence of chicken frame images. Based on a pre-constructed video analysis model, the state of the chicken flock is analyzed based on the sequence of chicken frame images to obtain a chicken flock state, which includes the overall distribution of the chicken flock, the activity level of the chickens, and the feather state of the chickens. Based on the overall distribution of the chicken flock, the activity level of the chickens, and the feather state of the chickens, the health level of the chicken flock is determined, and if the health level of the chicken flock is lower than a preset health level, a pre-warning scheme is generated based on a pre-constructed database. The pre-constructed digital twin model is updated in real time based on the health level of the chicken flock.

7. The method according to claim 6, wherein, After the pre-constructed digital twin model is updated in real time based on the health level of the chicken flock, the following operations are further included: Based on the updated digital twin model, an adaptive feeding strategy generation algorithm is used to generate updated feeding strategy parameters. The updated feeding strategy parameters are sent to the control center, so that the control center updates the composite feeding strategy based on the updated feeding strategy parameters to obtain feeding update control instructions, and regulates and controls the feeding line unit to feed the corresponding trough based on the feeding update control instructions.

8. A feed line adaptive precision feeding device for implementing the method of any one of claims 1 to 7, characterized in that, It includes: An information acquisition unit is configured to acquire multi-modal data collected by multi-modal sensors deployed in a chicken house and video sequences synchronously collected by wide-angle cameras deployed in the chicken house, the multi-modal data including body weight data, body temperature data, and environmental data. A graph construction unit is configured to identify the chicken behavior data based on the video sequences, construct a social network graph of the chicken flock based on the chicken behavior data using a graph neural network, and obtain individual indicators of each chicken based on the social network graph. A trend prediction unit is configured to fuse the multi-modal data, the chicken behavior data, and the individual indicators to obtain fusion data, input the fusion data into a multi-modal time series prediction model, and obtain growth trend prediction data of each chicken in a future preset time period, the growth trend prediction data including health trend, growth curve deviation risk, and potential disease risk. A strategy generation unit is configured to acquire real-time monitored feed consumption, generate a composite feeding strategy based on a multi-objective optimization algorithm for the growth trend prediction data of each chicken in the future preset time period and the feed consumption, and obtain the composite feeding strategy, the composite feeding strategy including target feeding amount, feeding time window, and feed formula ratio. The control feeding unit is configured to send the composite feeding strategy to the control center, so that the control center generates control instructions according to the composite feeding strategy, and regulates and controls the feeding line unit to feed the corresponding trough based on the control instructions.

9. A computer device, comprising: The computer program or the instructions realize the feeding line adaptive precision feeding method as claimed in any one of claims 1-7 when executed by a computer.

10. A computer program product comprising computer programs or instructions, characterized in that, The computer program or the instructions realize the feeding line adaptive precision feeding method as claimed in any one of claims 1-7 when executed by a computer.

Images