Rapid Determination Method and System for Key Nutrients in Laying Hen Feed Ingredients

Abstract

This invention provides a rapid method and system for determining key nutrients in laying hen feed ingredients, relating to the field of computer technology. The method includes: acquiring hyperspectral image data of the feed ingredient sample and laying hen production data; performing feature encoding based on the hyperspectral image data and laying hen production data to obtain a target sensitive feature set; mapping nutritional requirements based on the target sensitive feature set to obtain a stage-adaptive feature representation; simulating the digestion and absorption process based on the stage-adaptive feature representation to obtain bioavailable nutrient characteristics; optimizing the bioavailable nutrient characteristics to obtain key component weights that balance various physiological needs; and performing end-to-end prediction based on the key component weights, establishing an implicit functional relationship between the nutrient weights and laying hen production performance indicators to obtain quantitative values ​​of the key nutrients. This invention enables the measurement results to more accurately reflect the actual feeding value of the feed ingredients.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a method and system for rapid determination of key nutrients in laying hen feed ingredients. Background Technology

[0002] The detection and evaluation of key nutrients in laying hen feed ingredients is a core technological link in the field of precision nutrition and healthy breeding of laying hens. Its development has evolved from traditional chemical analysis to rapid instrument detection, and then to dynamic and systematic evaluation. In the early stage, it relied entirely on laboratory chemical analysis methods. Although the accuracy was recognized, it had inherent limitations such as long cycle, high cost, and sample destruction, making it difficult to support rapid decision-making in the breeding field. In order to improve efficiency, rapid detection methods such as near-infrared spectroscopy were introduced, which enabled the rapid prediction of some conventional components such as protein and moisture, marking an important shift from offline analysis to online detection. However, such technologies are essentially still the determination of static chemical components of raw materials, and have not broken through the scope of "raw materials themselves" to take into account the core receptor of laying hens and their complex physiological and metabolic laws.

[0003] In recent years, with the deepening demand for precision nutrition in the livestock industry, researchers have begun to try to integrate raw material data with animal production performance in hopes of establishing a more direct correlation. However, existing attempts are mostly limited to simple statistical regression or static models, failing to systematically characterize the dynamic effectiveness and bioavailability of nutrients in the entire chain from raw material intake, digestive tract digestion and absorption, liver metabolism and transformation to the final formation of eggs. In particular, there is a lack of fine quantification and synergistic optimization of specific egg-laying stages, physiological states and multi-objective health needs (such as egg production efficiency, immune maintenance and bone health) of laying hens. This leads to a frequent disconnect between test indicators and actual feeding benefits, forming a gap between data application and breeding results, which restricts further breakthroughs in precision feeding technology.

[0004] Based on the shortcomings of the existing technology, there is an urgent need for a rapid method and system for determining the key nutrients in laying hen feed ingredients. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid method and system for determining key nutrients in laying hen feed ingredients, thereby addressing the aforementioned problems. To achieve this objective, the technical solution adopted by this invention is as follows:

[0006] Firstly, this application provides a rapid method for determining key nutrient components in laying hen feed ingredients, including:

[0007] Acquire hyperspectral image data and egg production data of the raw material sample to be tested, wherein the egg production data includes egg production stage parameters and egg quality index data of the egg production hens associated with the historical feeding records of the raw material sample to be tested;

[0008] Feature encoding is performed based on the hyperspectral image data and the laying hen production data. An attention mechanism is constructed by using egg quality indicators as supervisory signals to screen spectral features related to egg production performance and obtain a target sensitive feature set.

[0009] Nutritional requirements are mapped based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of hens, the features are nonlinearly transformed to obtain stage-adaptive feature representations.

[0010] Based on the stage adaptation characteristics, the digestion and absorption process is simulated to obtain the characteristics of bioavailable nutrients.

[0011] Based on the characteristics of the bioactive nutrients, optimization processing is performed to obtain the weights of key components that balance various physiological needs.

[0012] End-to-end prediction is performed based on the weights of the key components. By establishing an implicit functional relationship between the weights of the nutrients and the performance indicators of laying hens, the quantitative values ​​of the key nutrients are obtained.

[0013] Secondly, this application also provides a rapid method and system for determining key nutrient components in laying hen feed ingredients, including:

[0014] The acquisition module is used to acquire hyperspectral image data and egg production data of the raw material sample to be tested. The egg production data includes egg production stage parameters and egg quality index data of the egg-laying hens associated with the historical feeding records of the raw material sample to be tested.

[0015] The encoding module is used to perform feature encoding based on the hyperspectral image data and the egg production data. By using egg quality indicators as supervision signals to construct an attention mechanism, it filters spectral features related to egg production performance to obtain a target sensitive feature set.

[0016] The mapping module is used to map nutritional requirements based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of laying hens, the features are nonlinearly transformed to obtain stage-adaptive feature representations.

[0017] The simulation module is used to simulate the digestion and absorption process based on the stage adaptation feature representation to obtain the characteristics of bioavailable nutrients.

[0018] The optimization module is used to optimize the process based on the characteristics of the bioactive nutrients to obtain the weights of key components that balance various physiological needs.

[0019] The prediction module is used to perform end-to-end prediction based on the weights of the key components. By establishing an implicit functional relationship between the weights of the nutrients and the performance indicators of laying hens, the quantitative values ​​of the key nutrients are obtained.

[0020] The beneficial effects of this invention are as follows:

[0021] This invention integrates hyperspectral image data with specific production data of laying hens, encodes features guided by egg production performance, and then establishes an end-to-end predictive model directly related to production performance through dynamic mapping of nutritional requirements, simulation of digestion and absorption processes, and optimization of multi-objective physiological requirements. This achieves a transformation from the determination of static chemical composition of raw materials to rapid and accurate assessment of the bioavailable nutritional components of laying hens, making the measurement results more accurately reflect the actual feeding value of raw materials. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart illustrating a rapid determination method for key nutrients in laying hen feed ingredients as described in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the structure of a rapid determination system for key nutrients in laying hen feed ingredients as described in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of a rapid determination device for key nutrients in laying hen feed ingredients, as described in an embodiment of the present invention.

[0026] The diagram is labeled as follows: 800, a rapid determination device for key nutrients in laying hen feed ingredients; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component; 901, acquisition module; 902, encoding module; 903, mapping module; 904, simulation module; 905, optimization module; 906, prediction module. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0029] Example 1

[0030] This embodiment provides a rapid method for determining key nutrients in laying hen feed ingredients.

[0031] See Figure 1 The figure shows that the method includes steps S100 to S600.

[0032] Step S100: Obtain hyperspectral image data and egg production data of the raw material sample to be tested. The egg production data includes egg production stage parameters and egg quality index data of the egg-laying hens associated with the historical feeding records of the raw material sample to be tested.

[0033] Understandably, hyperspectral image data is obtained by scanning raw material samples using a hyperspectral imaging system. It contains reflectance spectral information of the sample across hundreds of continuous narrow bands, accurately reflecting the spatial distribution and physical structure characteristics of chemical components such as proteins, fats, and fibers within the raw material. Egg production data, on the other hand, is a dynamic, multi-dimensional dataset, specifically including: 1) parameters related to the egg production stage, such as flock age, egg production rate curves, and the current egg production cycle (e.g., onset, peak, and late stages); 2) egg quality indicators linked to historical feeding records, such as egg weight, Haugh units, eggshell thickness, and yolk color—data directly reflecting feeding effectiveness. This data originates from the database records of farm production management systems (such as ERP) or is automatically collected by IoT devices. This step aims to correlate the static physicochemical properties of the raw materials with the dynamic production performance of the egg-laying hen population, laying a data foundation for subsequent evaluation models oriented towards actual production benefits.

[0034] Step S200: Based on hyperspectral image data and egg production data, feature encoding is performed. An attention mechanism is constructed by using egg quality indicators as supervisory signals to screen spectral features related to egg production performance and obtain a target sensitive feature set.

[0035] It should be noted that the purpose of this step is to screen out key spectral features that are truly indicative of egg production performance from a massive amount of raw features. Unlike traditional methods that select features solely based on the statistical relationship between spectra and chemical composition, this embodiment uses "egg quality," the final economic indicator, as a supervisory signal to guide feature selection. By constructing an attention mechanism, the model can automatically learn and assign greater weight to spectral bands or feature combinations that are more correlated with high-quality eggs (such as high Hardy units and suitable eggshell thickness). This filters out a large amount of redundant or interfering information that is chemically significant but has no significant impact on actual production benefits, allowing subsequent processing to focus on the feature subset most sensitive to breeding efficiency.

[0036] Step S300: Map nutritional requirements based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of laying hens, nonlinear transformation is performed on the features to obtain stage-adapted feature representations.

[0037] It is understandable that laying hens have significantly different nutritional requirements for energy, amino acids, calcium, and phosphorus at different laying stages (e.g., maintaining high production during peak periods and maintaining eggshell quality in later stages). This treatment does not involve simple linear adjustments to the ratios, but rather constructs a nonlinear transformation model to deeply convert the spectral characteristics obtained in the previous step according to the ideal nutritional requirement curve corresponding to the specific physiological stage of the laying hen. Essentially, it maps general raw material chemical composition information into customized nutritional supply characteristics that match the physiological goals of a specific laying stage, making the feature representation inherently marked with the nutritional requirements of that stage.

[0038] Step S400: Simulate the digestion and absorption process based on the stage adaptation feature representation to obtain the characteristics of bioavailable nutrients;

[0039] It should be noted that this step incorporates the principles of laying hen digestive physiology, simulating a series of physical and biochemical processes at the algorithmic level, including nutrient grinding in the gizzard, intestinal enzymatic digestion, and contact with absorption surfaces. By constructing a digestive kinetic model, the release rate and absorption site of nutrients are simulated, and factors affecting bioavailability, such as first-pass metabolism in the liver, are corrected for. This transforms the theoretical nutrient characteristics into bioavailable nutrient characteristics that more closely approximate actual in vivo utilization, enhancing the correlation between the measurement results and the actual physiological responses of animals.

[0040] Step S500: Optimize the process based on the characteristics of bioavailable nutrients to obtain the weights of key components that balance various physiological needs;

[0041] Understandably, the nutritional needs of laying hens constitute a multi-objective balance system, with competing objectives such as high egg production efficiency, robust immune function, and good bone health. The core of this step is to address the trade-offs between these objectives. By constructing an optimization model, we simulate the internal nutrient allocation strategies of the laying hen under different physiological states, aiming to find a weighted allocation scheme for key components. This scheme does not pursue the maximization of a single objective (such as egg production rate), but rather seeks, in the sense of Pareto optimality, the best balance point that simultaneously and effectively meets multiple physiological needs such as egg production, immunity, and bone health, thereby making the final nutritional recommendations more holistic and considering animal welfare.

[0042] Step S600: Perform end-to-end prediction based on the weights of key components. By establishing an implicit functional relationship between the weights of nutrients and the performance indicators of laying hens, the quantitative values ​​of key nutrients are obtained.

[0043] It should be noted that this step is the final prediction and integration step. Its core idea is to establish an end-to-end, learnable implicit functional relationship between the "balanced nutrient weights" and "predictable production performance indicators." This method avoids the error accumulation caused by traditionally relying on multiple explicit, isolated empirical formulas linked together. By placing nutrient weights and laying hen production performance indicators (such as egg production rate and egg quality) under a unified mathematical framework for bidirectional constraint optimization (i.e., nutrient intake affects production performance, and production demand also inversely regulates nutrient utilization), the final quantitative value of the nutrient components is an optimal solution at the system level, ensuring that the measured values ​​can most effectively support the achievement of expected production targets.

[0044] Further, step S200 includes steps S210 to S230.

[0045] Step S210: Extract spectral features from hyperspectral image data, and extract multi-scale spatial-spectral features through convolution operation to obtain an initial high-dimensional feature map;

[0046] Step S220: Perform time-series dynamic weighting based on the initial high-dimensional feature map and the egg-laying stage parameters of the laying hen. By using the egg-laying stage parameters as modulation factors to dynamically scale the feature map, a stage-sensitive feature representation is obtained.

[0047] Step S230: Based on the stage-sensitive feature representation and egg quality index data, target-oriented feature screening is performed. By constructing an attention weight matrix with egg quality indicators as supervision signals, feature combinations related to egg production performance are screened to obtain the target sensitive feature set.

[0048] Specifically, in the refinement of feature encoding, steps S210 to S230 achieve a progressive extraction from the original hyperspectral image to the target sensitive feature set. First, in S210, spectral features are extracted from the hyperspectral image through convolution operations. Multi-scale convolution kernels (such as small-scale kernels capturing subtle textures and large-scale kernels capturing macroscopic structures) are used to slide and compute in both spatial and spectral dimensions, generating an initial high-dimensional feature map. This process essentially transforms pixel-level spectral reflectance data into a more discriminative feature representation. The weights of the convolution kernels are optimized through model training to highlight local patterns related to nutritional components in the raw material (such as the spectral features of protein-rich regions). Preferably, the input hyperspectral image H has dimensions X×Y×B, where X and Y are spatial dimensions, and B is the number of bands; for each scale s, the convolution kernel W... s Dimension K s ×K s ×B, where K s For kernel size (e.g., K) s =3 For small scales, K s =5 for large scales); Output feature map F s The dimension is XxY, and the computation is performed through discrete convolution, where the convolution kernel weights W s The training data is optimized to maximize the ability of the features to represent the chemical composition of the raw materials. The value at each position (i,j) of the feature map is represented as follows:

[0049] ;

[0050] In the formula, F s (i,j) represents the feature value of the output feature map at spatial location (i,j) at the s-th scale. This value is calculated through convolution and reflects the multi-scale spectral features of the input image at that location. H(i+u-1,j+v-1,b) represents the spectral reflectance value of the input hyperspectral image at spatial location (i+u-1,j+v-1,b) and the b-th band, where i and j are the spatial indices of the output feature map, and u and v are the relative position indices within the convolution kernel. W s (u,v,b) represents the weight value of the convolution kernel at the s-th scale at position (u,v) and in the b-th band. K s This represents the spatial dimensions (height and width) of the s-th scale convolution kernel. B represents the total number of bands in the hyperspectral image, i.e., the length of the spectral dimension.

[0051] Next, in S220, laying stage parameters of the laying hen (such as age or laying rate curve) are introduced as dynamic modulation factors. A learnable weight function is used to perform channel-level scaling of the feature map. For example, during peak laying, the weights of features related to energy metabolism are automatically increased, while calcium and phosphorus-related features are strengthened in the later stages of laying. This allows the feature representation to adapt to the temporal changes of physiological stages, resulting in stage-sensitive feature representations. This process overcomes the limitation of static consistency of features in traditional methods and achieves alignment with the physiological dynamics of the laying hen. Preferably, the input feature map F has dimensions C×H×W (where C is the number of channels, obtained from the concatenation of outputs from S210), and the laying stage parameter P is a scalar or vector (such as the laying stage index). The modulation function M(P) outputs a weight vector dimension, implemented through a fully connected layer and an activation function, for example, M(P) = σ(W). p P+b p The weighted feature map G is obtained by channel-level multiplication, where W_p and b_p are learning parameters, and σ is the Sigmoid function. Each channel is multiplied by its corresponding weight. The temporal dynamic weighting formula is expressed as:

[0052] ;

[0053] In the formula, G(c,h,w) represents the value of the stage-sensitive feature representation obtained after temporal dynamic weighting at channel c and spatial position (h,w). F(c,h,w) represents the value of the input feature map at channel c and spatial position (h,w). M(P) c This represents the weight value of the c-th channel calculated by the modulation function M based on the laying stage parameter P. P represents the laying stage parameters of the laying hen, such as laying age, laying rate, and other quantitative indicators. c is the index of the feature channel; h and w are the spatial height and width indices of the feature map.

[0054] Finally, in S230, using egg quality indicators (such as Haugh units and eggshell thickness) as supervisory signals, an attention weight matrix is ​​constructed to calculate the contribution of each feature channel to egg quality. A soft attention mechanism is then used to select high-contribution feature combinations, forming a target sensitive feature set. This step ensures that feature selection is directly guided by end-production benefits, avoiding interference from irrelevant features and thus improving the accuracy of subsequent nutritional assessments. Preferably, the input is a weighted feature map G, and the egg quality indicator Q is a vector (such as Haugh units and eggshell thickness). First, global average pooling is performed on G to obtain the feature vector g with dimension C, i.e.

[0055] ;

[0056] Then, the attention weight α and dimension C are calculated using a query-key attention mechanism, where the query is Q, the key is g, and the scoring function is... W aThe learning matrix is ​​used; the attention weights are normalized using softmax. Final weighted feature vector As the target-sensitive feature set T0, the target-oriented attention selection formula is integrated as follows:

[0057] ;

[0058] In the formula, T represents the transpose of the matrix; g j This represents the feature vector of the j-th feature channel after global average pooling.

[0059] Further, step S300 includes steps S310 to S330.

[0060] Step S310: Construct a nutritional requirement benchmark based on the target sensitive feature set, and obtain a dynamic nutritional requirement benchmark vector by fitting the ideal amino acid pattern curves at different egg production stages.

[0061] Step S320: Perform physiological metabolism simulation based on dynamic nutritional demand baseline vector, and obtain the theoretical available nutrient spectrum by establishing a nutrient conversion efficiency model with liver metabolism as the core.

[0062] Step S330: Based on the theoretically available nutrient spectrum and actual production performance data, the difference compensation is performed. By backpropagating the implicit relationship between egg quality indicators and nutrient utilization rate, the nonlinear transformation process of feature mapping is optimized to obtain the stage-adapted feature representation.

[0063] It should be noted that in the detailed processing of nutrient requirement mapping, steps S310 to S330 realize the transformation process from target-sensitive features to stage-adaptive feature representation. First, in S310, based on the target-sensitive feature set and the laying stage parameters of the laying hen, a dynamic nutrient requirement baseline vector is constructed by fitting the ideal amino acid pattern curves of different laying stages. This process is achieved by establishing a nonlinear mapping function between the stage parameters and the ideal amino acid ratio. The laying stage parameters are used as input, and through fully connected layers and activation function transformations, a baseline vector that conforms to the nutrient requirements of that stage is output. This vector represents the ideal ratio of various nutrients required by the laying hen at a specific laying stage. Next, in S320, using the above dynamic nutrient requirement baseline vector as input, a nutrient conversion efficiency model with liver metabolism as the core is constructed to simulate physiological metabolism. This model uses a multi-layer neural network structure to simulate the conversion path of nutrients in the liver, where each layer corresponds to a metabolic link. The conversion efficiency between different nutrients is represented by a weight matrix, and finally, a theoretically available nutrient spectrum is output. This spectrum reflects the actual bioavailability of nutrients after metabolism by the body. Finally, in S330, the theoretical available nutrient spectrum is compared with the actual production performance data, and the feature mapping process is optimized by establishing a difference compensation mechanism. This mechanism uses the implicit relationship between egg quality indicators and nutrient utilization rate as a supervision signal, and uses the backpropagation algorithm to calculate the gradient difference between theoretical and actual values. By adjusting the parameters of the feature mapping network, the output features are better matched with the actual production performance, and finally the stage-adapted feature representation is obtained.

[0064] Further, step S400 includes steps S410 to S430.

[0065] Step S410: Model the digestive tract structure based on the stage adaptation feature representation, and construct a dynamic digestive network with gizzard grinding efficiency and intestinal absorption surface area as parameters by combining the physiological parameters of the laying hen's digestive tract, so as to obtain the physical digestion simulation results of the digestive tract.

[0066] Step S420: Perform digestion kinetics simulation based on the results of physical digestion simulation of the digestive tract. By establishing a differential equation model based on the chyme flow rate and enzymatic hydrolysis kinetics, simulate the stepwise release process of nutrients in the digestive tract and obtain the theoretical nutrient release spectrum.

[0067] Step S430: Perform bioavailability correction processing based on the theoretical nutrient release spectrum. By using the liver first-pass effect and intestinal mucosal transport efficiency as constraints, the nutrient absorption rate is dynamically corrected to obtain the bioavailable nutrient characteristics.

[0068] Specifically, firstly, in S410, the digestive tract structure is modeled based on stage-adaptive feature representations, and a dynamic digestive network is constructed by combining digestive tract physiological parameters specific to laying hens. This network uses gizzard grinding efficiency and intestinal absorptive surface area as core parameters, and simulates the physical breakage process of feed in the gizzard and the absorption environment in the intestine through a multilayer perceptron. The gizzard grinding efficiency is determined by the functional relationship between feed hardness and gizzard muscle activity, while the intestinal absorptive surface area dynamically changes with intestinal peristalsis. Finally, the physical digestion simulation results are output, quantifying the initial degree of feed breakage and accessible surface area. The digestive tract structure modeling process is represented as follows:

[0069] ;

[0070] In the formula, D p This indicates the results of a physical digestion simulation of the digestive tract; F a The stage-adaptive feature representation is represented by Θ; Θ represents the digestive tract physiological parameter vector, including parameters such as gizzard grinding efficiency and intestinal absorptive surface area; W d The weight matrix of the digestion network represents the mapping relationship learned from the physical digestion process; b d This is the paranoia vector, used to adjust the baseline level of the output; This is a vector concatenation operation; Ψ is the Sigmoid activation function, ensuring that the output value is within a reasonable range. This represents the mathematical process of modeling the entire digestive tract structure.

[0071] Next, in S420, digestive kinetics simulation is performed based on the results of physical digestion simulation of the digestive tract. Differential equations are established to describe the chyme flow rate and enzymatic hydrolysis kinetics. These equations simulate the stepwise release of nutrients in different segments of the digestive tract (provitative stomach, gizzard, small intestine). The chyme flow rate is determined by the peristaltic frequency and the viscosity of the contents, while the enzymatic hydrolysis kinetics follow the reaction rate described by the Michaelis-Menten equation. Finally, a theoretical nutrient release profile is obtained, reflecting the release sequence and magnitude of nutrients under ideal conditions. The digestive kinetics differential equation is:

[0072] ;

[0073] In the formula, N m (t) represents the nutrient concentration in the m-th digestive tract segment at time t; This indicates the rate of change of nutrient concentration over time; v flow The rate of chyme flow is determined by the frequency of peristalsis in the digestive tract and the viscosity of the contents; v enz K represents the maximum rate of enzymatic hydrolysis, reflecting the activity level of the digestive enzyme; m N is the Michaelis constant, representing the reciprocal of the enzyme's affinity for its substrate; m-1(t) represents the nutrient concentration in the upstream segment, which serves as the input for the current segment.

[0074] Finally, in S430, the theoretical nutrient release profile was bioavailability corrected, with the hepatic first-pass effect and intestinal mucosal transport efficiency as constraints. The hepatic first-pass effect was simulated by a decay function to represent the metabolic loss of nutrients in the portal circulation, while the intestinal mucosal transport efficiency was modeled based on a carrier protein saturation kinetics model. Through this dual correction of the absorption rate, the bioavailable nutrient characteristics that better reflect actual in vivo utilization were obtained. The correction formula is expressed as:

[0075] ;

[0076] In the formula, F b The characteristics of bioavailable nutrients indicate the actual amount of nutrients that can be utilized by the body; U r To theoretically release the nutrient spectrum; A intestinal This is a vector of intestinal mucosal transport efficiency, where each element represents the absorption rate of a specific nutrient; Represents element-wise multiplication (Hadamard product); L first-pass T is the first-pass loss vector, representing the proportion of nutrients lost during liver metabolism; max K represents the maximum transfer rate. t is the transport half-saturation constant.

[0077] Further, step S500 includes steps S510 to S530.

[0078] Step S510: Based on the characteristics of bioavailable nutrients and the multi-objective requirements of laying hens, demand conflict modeling is performed. By constructing a multi-objective optimization surface with egg production efficiency, immune status and bone health as vertices, a competitive relationship model among the various requirements is established.

[0079] Step S520: Perform dynamic weight optimization based on the competition relationship model. By simulating the nutrient allocation strategy of laying hens under different physiological states, find the optimal solution set that meets the real-time requirements on the Pareto front.

[0080] Step S530: Adaptive adjustment is performed based on the optimal solution set. By using production performance indicators from historical feeding data as feedback signals, gradient optimization is performed on the weight allocation to obtain the key component weights that balance various physiological needs.

[0081] Understandably, steps S510 to S530 realize the decision-making process from the characteristics of bioavailable nutrients to the weighting of key components that balance various physiological needs. First, in S510, demand conflict modeling is performed based on the characteristics of bioavailable nutrients and the multi-objective demand parameters of laying hens, by constructing a multi-objective optimization surface with egg production efficiency, immune status, and bone health as vertices. This process adopts multi-objective optimization theory, transforming each physiological need into an independent objective function. The egg production efficiency objective function mainly focuses on the efficiency of energy and amino acid utilization, the immune status objective function focuses on the sufficient supply of antioxidants and immune-enhancing factors, and the bone health objective function focuses on the calcium-phosphorus ratio and trace element balance. The three objective functions have a natural competitive relationship in nutrient allocation, and this trade-off relationship is characterized by constructing a Pareto front. Next, in S520, dynamic weight optimization is performed based on the aforementioned competitive relationship model, simulating the nutrient allocation strategy of laying hens under different physiological states. This process employs an adaptive weight adjustment mechanism, dynamically adjusting the relative importance of each objective based on real-time physiological signals of the laying hens (such as changes in egg production rate, body temperature fluctuations, and behavioral characteristics). For example, higher weights are assigned to egg production efficiency during peak laying periods, and the weight of immune status is increased under stress. Through efficient search in a multi-dimensional objective space, the optimal solution set located on the Pareto front is found. Finally, in S530, the optimal solution set is adaptively adjusted, using production performance indicators from historical feeding data as feedback signals. This process establishes a closed-loop optimization system. By comparing the difference between the theoretical optimal solution and the actual feeding effect, the gradient of the loss function is calculated, and the backpropagation algorithm is used to fine-tune the weight allocation strategy, making the theoretical model better fit the actual production situation, ultimately obtaining key component weights that are both consistent with nutritional principles and validated in practice.

[0082] Further, step S600 includes steps S610 to S630.

[0083] Step S610: Construct a prediction function based on the weights of key components. By embedding the rate of change in body weight and physiological signals of feather quality status as latent variables into the neural network structure, a nutrient-production performance mapping function is established.

[0084] Step S620: Perform function optimization processing based on the nutrient-production performance mapping function. By using the production performance fluctuations at different physiological stages as training constraints, the time-series difference method is used to dynamically calibrate the function parameters to obtain the calibrated mapping function.

[0085] Step S630: Perform bidirectional prediction based on the calibrated mapping function. By simultaneously considering the positive impact of nutrient intake on production performance and the negative regulation of nutrient utilization by production demand, the quantitative values ​​of key nutrients are obtained.

[0086] Specifically, in S610, a prediction function is constructed based on the weights of key components by embedding dynamic physiological signals such as weight change rate and feather quality as latent variables into the neural network structure. This process employs a deep neural network architecture, combining static nutrient weights with dynamic physiological monitoring signals. The weight change rate reflects the energy balance of the laying hen, while feather quality indirectly indicates the metabolic status of protein and trace elements. By encoding these two types of signals as latent variables and integrating them into the network's hidden layers, a nutrient-production performance mapping function that simultaneously considers nutrient supply and physiological state is constructed. Next, in S620, function optimization is performed based on the initially constructed nutrient-production performance mapping function, using production performance fluctuations at different physiological stages as training constraints. This process pays particular attention to key turning points in the laying hen's physiological cycle, such as the transition from peak egg production to later stages and molting. The temporal difference method is used to compare the prediction errors of adjacent time points, dynamically calibrating the function parameters so that the mapping function can adapt to continuous changes in the laying hen's physiological state, ensuring the prediction model maintains accuracy at different physiological stages. The calibrated mapping function is expressed as:

[0087] ;

[0088] In the formula, f calibrated (W k H p ,t) is the calibrated nutrient-production performance mapping function; W k H is the key component weight vector; p f is a vector of latent variables representing physiological signals; map This represents the initial nutrient-production performance mapping function structure; Θ map ΔΘ represents the set of parameters for the initial mapping function. map (t) represents the parameter adjustment amount calculated by the time-series difference method at time t.

[0089] Finally, in S630, bidirectional prediction is performed based on the calibrated mapping function. This innovative method breaks through the limitations of traditional unidirectional prediction by simultaneously considering the positive impact of nutrient intake on production performance and the reverse regulation of nutrient utilization by production demand. It establishes a bidirectional coupling relationship between nutrients and performance. The positive path simulates how nutrients affect production performance through metabolic pathways, while the reverse path describes how production demand (such as eggshell formation and feather renewal) regulates the body's absorption and utilization efficiency of nutrients. By solving the optimal solution of this bidirectional constraint system, quantitative values ​​of key nutrients that meet both nutritional supply and physiological needs are obtained.

[0090] Based on the above two-way prediction process, taking soybean meal samples as an example, the quantitative values ​​of key nutrients obtained through the complete rapid determination method are shown in the table below:

[0091] Table 1. Quantitative determination results of key nutrients in soybean meal samples

[0092]

[0093] The example demonstrates the measurement results of soybean meal feed for laying hens during their peak laying period. Compared with traditional chemical analysis methods, the measurement values ​​provided by this invention have undergone digestion and absorption process simulation and bioavailability correction (e.g., crude protein was corrected from 45.2% to 43.8% of actual usable protein). Through multi-objective optimization, weight coefficients suitable for the peak laying period were assigned to different nutrients, with higher weights for protein and amino acids. These measurement values ​​simultaneously meet the dual constraints of nutrient supply efficiency and production performance requirements, conforming to nutritional requirements and accurately predicting actual feeding effects. They can be directly used in laying hen feed formulation design, effectively improving feed utilization efficiency and production performance.

[0094] Example 2:

[0095] like Figure 2 As shown, this embodiment provides a rapid determination system for key nutrients in laying hen feed ingredients. The system includes:

[0096] The acquisition module 901 is used to acquire hyperspectral image data and egg production data of the raw material sample to be tested. The egg production data includes egg production stage parameters and egg quality index data of the egg-laying hens associated with the historical feeding records of the raw material sample to be tested.

[0097] Encoding module 902 is used to perform feature encoding based on hyperspectral image data and egg production data. By using egg quality indicators as supervision signals to construct an attention mechanism, it filters spectral features related to egg production performance to obtain a target sensitive feature set.

[0098] The mapping module 903 is used to map nutritional requirements based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of laying hens, the features are nonlinearly transformed to obtain stage-adapted feature representations.

[0099] Simulation module 904 is used to simulate the digestion and absorption process based on the stage adaptation feature representation to obtain the characteristics of bioavailable nutrients.

[0100] The optimization module 905 is used to optimize the process based on the characteristics of bioavailable nutrients to obtain the weights of key components that balance various physiological needs.

[0101] The prediction module 906 is used to perform end-to-end prediction based on the weights of key components. By establishing an implicit functional relationship between the weights of nutrients and the performance indicators of laying hens, the quantitative values ​​of key nutrients are obtained.

[0102] In this specific embodiment of the present application, the encoding module includes:

[0103] The first coding unit is used to extract spectral features from hyperspectral image data and extract multi-scale spatial-spectral features through convolution operations to obtain an initial high-dimensional feature map.

[0104] The second coding unit is used to perform time-series dynamic weighting based on the initial high-dimensional feature map and the egg-laying stage parameters of the laying hen. By using the egg-laying stage parameters as modulation factors to dynamically scale the feature map, a stage-sensitive feature representation is obtained.

[0105] The third coding unit is used to perform target-oriented feature screening based on stage-sensitive feature representation and egg quality index data. By constructing an attention weight matrix using egg quality indicators as supervision signals, it screens feature combinations related to egg production performance to obtain the target-sensitive feature set.

[0106] In this specific embodiment of the present application, the mapping module includes:

[0107] The first mapping unit is used to construct a nutritional requirement benchmark based on the target sensitive feature set. By fitting the ideal amino acid pattern curves at different egg-laying stages, a dynamic nutritional requirement benchmark vector is obtained.

[0108] The second mapping unit is used to perform physiological metabolism simulation based on the dynamic nutritional demand baseline vector. By establishing a nutrient conversion efficiency model with liver metabolism as the core, the theoretical available nutrient spectrum is obtained.

[0109] The third mapping unit is used to compensate for the difference between the theoretically available nutrient spectrum and the actual production performance data. By backpropagating the implicit relationship between egg quality indicators and nutrient utilization rate, it optimizes the nonlinear transformation process of feature mapping and obtains the stage-adapted feature representation.

[0110] In this specific embodiment of the present application, the simulation module includes:

[0111] The first simulation unit is used to model the digestive tract structure based on the stage adaptation feature representation, and to construct a dynamic digestive network with gizzard grinding efficiency and intestinal absorptive surface area as parameters by combining the physiological parameters of the laying hen's digestive tract, so as to obtain the physical digestion simulation results of the digestive tract.

[0112] The second simulation unit is used to perform digestive kinetic simulation based on the results of physical digestion simulation of the digestive tract. By establishing a differential equation model based on the chyme flow rate and enzymatic hydrolysis kinetics, it simulates the stepwise release process of nutrients in the digestive tract and obtains the theoretical nutrient release spectrum.

[0113] The third simulation unit is used to perform bioavailability correction based on the theoretical nutrient release profile. By using the liver first-pass effect and intestinal mucosal transport efficiency as constraints, the nutrient absorption rate is dynamically corrected to obtain the bioavailable nutrient characteristics.

[0114] In this specific embodiment of the present application, the optimization module includes:

[0115] The first optimization unit is used to model the demand conflict based on the characteristics of bioavailable nutrients and the multi-objective demand parameters of laying hens. It establishes a model of the competitive relationship between various demands by constructing a multi-objective optimization surface with egg production efficiency, immune status and bone health as vertices.

[0116] The second optimization unit is used to perform dynamic weight optimization based on the competition relationship model. By simulating the nutrient allocation strategy of laying hens under different physiological states, it seeks the optimal solution set that meets real-time requirements on the Pareto front.

[0117] The third optimization unit is used to make adaptive adjustments based on the optimal solution set. By using production performance indicators from historical feeding data as feedback signals, it performs gradient optimization on the weight allocation to obtain the key component weights that balance various physiological needs.

[0118] In this specific embodiment of the present application, the prediction module includes:

[0119] The first prediction unit is used to construct a prediction function based on the weights of key components. It establishes a nutrient-production performance mapping function by embedding the rate of change in body weight and physiological signals of feather quality as latent variables into the neural network structure.

[0120] The second prediction unit is used to perform function optimization based on the nutrient-production performance mapping function. By using the production performance fluctuations at different physiological stages as training constraints, the function parameters are dynamically calibrated using the temporal difference method to obtain the calibrated mapping function.

[0121] The third prediction unit is used to make bidirectional predictions based on the calibrated mapping function. By simultaneously considering the positive impact of nutrient intake on production performance and the negative regulation of nutrient utilization by production demand, it solves for the quantitative values ​​of key nutrients.

[0122] Example 3

[0123] Corresponding to the above method embodiments, this embodiment also provides a rapid determination device for key nutrients in laying hen feed ingredients. The rapid determination device for key nutrients in laying hen feed ingredients described below and the rapid determination method for key nutrients in laying hen feed ingredients described above can be referred to in correspondence.

[0124] Figure 3This is a block diagram illustrating a rapid determination device 800 for key nutrient components of laying hen feed ingredients, according to an exemplary embodiment. Figure 3 As shown, the rapid determination device 800 for key nutrients in laying hen feed ingredients may include: a processor 801 and a memory 802. The rapid determination device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0125] The processor 801 controls the overall operation of the rapid determination device 800 for key nutrients in laying hen feed ingredients to complete all or part of the steps in the aforementioned rapid determination method for key nutrients in laying hen feed ingredients. The memory 802 stores various types of data to support the operation of the rapid determination device 800 for key nutrients in laying hen feed ingredients. This data may include, for example, instructions for any application or method operating on the rapid determination device 800 for key nutrients in laying hen feed ingredients, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as keyboards, mice, and buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the rapid determination device 800 for key nutrients in laying hen feed and other devices. Wireless communication includes Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0126] In an exemplary embodiment, a rapid determination device 800 for key nutrients in laying hen feed ingredients may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned rapid determination method for key nutrients in laying hen feed ingredients.

[0127] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, these program instructions implement the steps of the rapid determination method for key nutrients in laying hen feed ingredients described above. For example, the computer-readable storage medium may be the memory 802 including the program instructions, which may be executed by the processor 801 of a rapid determination device 800 for key nutrients in laying hen feed ingredients to complete the rapid determination method for key nutrients in laying hen feed ingredients described above.

[0128] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for rapid determination of key nutrient components of a layer feed raw material, characterized in that, include: Acquire hyperspectral image data and egg production data of the raw material sample to be tested, wherein the egg production data includes egg production stage parameters and egg quality index data of the egg production hens associated with the historical feeding records of the raw material sample to be tested; Feature encoding is performed based on the hyperspectral image data and the laying hen production data. An attention mechanism is constructed by using egg quality indicators as supervisory signals to screen spectral features related to egg production performance and obtain a target sensitive feature set. Nutritional requirements are mapped based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of hens, the features are nonlinearly transformed to obtain stage-adaptive feature representations. Based on the stage adaptation characteristics, the digestion and absorption process is simulated to obtain the characteristics of bioavailable nutrients. Based on the characteristics of the bioactive nutrients, optimization processing is performed to obtain the weights of key components that balance various physiological needs. End-to-end prediction is performed based on the weights of the key components. By establishing an implicit functional relationship between the weights of the nutrients and the performance indicators of laying hens, the quantitative values ​​of the key nutrients are obtained.

2. The rapid determination method for key nutrients in laying hen feed ingredients according to claim 1, characterized in that, Feature encoding is performed based on the hyperspectral image data and the egg production data, including: Spectral features are extracted from the hyperspectral image data, and multi-scale spatial-spectral features are extracted through convolution operations to obtain an initial high-dimensional feature map. Based on the initial high-dimensional feature map and the egg-laying stage parameters of the laying hen, a time-series dynamic weighting is performed, and the feature map is dynamically scaled by using the egg-laying stage parameters as a modulation factor to obtain a stage-sensitive feature representation; Based on the stage-sensitive feature representation and the egg quality index data, target-oriented feature screening is performed. By constructing an attention weight matrix using egg quality indicators as supervision signals, feature combinations related to egg production performance are screened to obtain the target sensitive feature set.

3. The rapid determination method for key nutrients in laying hen feed ingredients according to claim 1, characterized in that, Nutritional requirement mapping based on the target sensitive feature set includes: Nutritional requirement benchmarks are constructed based on the target sensitive feature set, and dynamic nutritional requirement benchmark vectors are obtained by fitting the ideal amino acid pattern curves at different egg production stages. Physiological metabolism simulation is performed based on the dynamic nutritional requirement baseline vector. By establishing a nutrient conversion efficiency model with liver metabolism as the core, the theoretical available nutrient spectrum is obtained. According to the theory, the difference between nutrient profile and actual production performance data can be compensated. By backpropagating the implicit relationship between egg quality indicators and nutrient utilization rate, the nonlinear transformation process of feature mapping can be optimized to obtain stage-adaptive feature representation.

4. The rapid determination method for key nutrients in laying hen feed ingredients according to claim 1, characterized in that, The digestion and absorption process is simulated based on the stage adaptation feature representation, including: Based on the stage adaptation feature representation, the digestive tract structure is modeled, and a dynamic digestive network with gizzard grinding efficiency and intestinal absorptive surface area as parameters is constructed in combination with the physiological parameters of the laying hen's digestive tract to obtain the physical digestion simulation results of the digestive tract. Based on the results of the physical digestion simulation of the digestive tract, digestive kinetics simulation was performed. By establishing a differential equation model based on the chyme flow rate and enzymatic hydrolysis kinetics, the stepwise release process of nutrients in the digestive tract was simulated, and the theoretical nutrient release spectrum was obtained. Based on the theoretical release of nutrient profiles, bioavailability correction was performed. By using the liver first-pass effect and intestinal mucosal transport efficiency as constraints, the nutrient absorption rate was dynamically corrected to obtain the bioavailable nutrient characteristics.

5. The rapid determination method for key nutrient components of laying hen feed ingredients according to claim 1, characterized in that, Optimization processing based on the characteristics of the bioactive nutrients includes: Based on the characteristics of the bioavailable nutrients and the multi-objective requirements of laying hens, a demand conflict model is constructed. By building a multi-objective optimization surface with egg production efficiency, immune status and bone health as vertices, a competitive relationship model among the various requirements is established. Based on the aforementioned competitive relationship model, dynamic weight optimization is performed. By simulating the nutrient allocation strategy of laying hens under different physiological states, the optimal solution set that meets real-time requirements is found on the Pareto front. Adaptive adjustments are made based on the optimal solution set. By using production performance indicators from historical feeding data as feedback signals, gradient optimization is performed on the weight allocation to obtain the key component weights that balance various physiological needs.

6. A rapid determination system for key nutrient components in laying hen feed ingredients, characterized in that, include: The acquisition module is used to acquire hyperspectral image data and egg production data of the raw material sample to be tested. The egg production data includes egg production stage parameters and egg quality index data of the egg-laying hens associated with the historical feeding records of the raw material sample to be tested. The encoding module is used to perform feature encoding based on the hyperspectral image data and the egg production data. By using egg quality indicators as supervision signals to construct an attention mechanism, it filters spectral features related to egg production performance to obtain a target sensitive feature set. The mapping module is used to map nutritional requirements based on the target sensitive feature set. By establishing a nutritional requirement model for different egg-laying stages of laying hens, the features are nonlinearly transformed to obtain stage-adaptive feature representations. The simulation module is used to simulate the digestion and absorption process based on the stage adaptation feature representation to obtain the characteristics of bioavailable nutrients. The optimization module is used to optimize the process based on the characteristics of the bioactive nutrients to obtain the weights of key components that balance various physiological needs. The prediction module is used to perform end-to-end prediction based on the weights of the key components. By establishing an implicit functional relationship between the weights of the nutrients and the performance indicators of laying hens, the quantitative values ​​of the key nutrients are obtained.

7. The rapid determination system for key nutrients in laying hen feed ingredients according to claim 6, characterized in that, The encoding module includes: The first encoding unit is used to extract spectral features based on the hyperspectral image data, and extract multi-scale spatial-spectral features through convolution operations to obtain an initial high-dimensional feature map. The second encoding unit is used to perform time-series dynamic weighting based on the initial high-dimensional feature map and the egg-laying stage parameters of the laying hen, and to dynamically scale the feature map by using the egg-laying stage parameters as modulation factors to obtain a stage-sensitive feature representation. The third encoding unit is used to perform target-oriented feature screening based on the stage-sensitive feature representation and the egg quality index data. By constructing an attention weight matrix using egg quality indicators as a supervision signal, it screens feature combinations related to egg production performance to obtain the target sensitive feature set.

8. The rapid determination system for key nutrient components of laying hen feed ingredients according to claim 6, characterized in that, The mapping module includes: The first mapping unit is used to construct a nutritional requirement benchmark based on the target sensitive feature set and obtain a dynamic nutritional requirement benchmark vector by fitting the ideal amino acid pattern curves at different egg-laying stages. The second mapping unit is used to perform physiological metabolism simulation based on the dynamic nutritional demand benchmark vector. By establishing a nutrient conversion efficiency model with liver metabolism as the core, the theoretically available nutrient spectrum is obtained. The third mapping unit is used to compensate for the difference between the theoretically available nutrient spectrum and the actual production performance data. By backpropagating the implicit relationship between egg quality indicators and nutrient utilization rate, it optimizes the nonlinear transformation process of feature mapping and obtains the stage-adapted feature representation.

9. The rapid determination system for key nutrient components of laying hen feed ingredients according to claim 6, characterized in that, The simulation module includes: The first simulation unit is used to model the digestive tract structure based on the stage adaptation feature representation, and to construct a dynamic digestive network with gizzard grinding efficiency and intestinal absorptive surface area as parameters by combining the physiological parameters of the laying hen's digestive tract, so as to obtain the physical digestion simulation results of the digestive tract. The second simulation unit is used to perform digestive kinetics simulation based on the physical digestion simulation results of the digestive tract. By establishing a differential equation model based on the chyme flow rate and enzymatic hydrolysis kinetics, it simulates the stepwise release process of nutrients in the digestive tract and obtains the theoretical nutrient release spectrum. The third simulation unit is used to perform bioavailability correction processing based on the theoretical nutrient release spectrum. By using the liver first-pass effect and intestinal mucosal transport efficiency as constraints, the nutrient absorption rate is dynamically corrected to obtain the bioavailable nutrient characteristics.

10. The rapid determination system for key nutrients in laying hen feed ingredients according to claim 6, characterized in that, The optimization module includes: The first optimization unit is used to model the demand conflict based on the characteristics of the bioavailable nutrients and the multi-objective demand parameters of laying hens. By constructing a multi-objective optimization surface with egg production efficiency, immune status and bone health as vertices, a model of the competitive relationship between various demands is established. The second optimization unit is used to perform dynamic weight optimization based on the competition relationship model, and to find the optimal solution set that meets the real-time requirements on the Pareto front by simulating the nutrient allocation strategy of laying hens under different physiological states. The third optimization unit is used to make adaptive adjustments based on the optimal solution set. By using production performance indicators from historical feeding data as feedback signals, it performs gradient optimization on the weight allocation to obtain the key component weights that balance various physiological needs.

Images