Mutton quality classification method and system based on multi-source data analysis

Abstract

The invention relates to the technical field of quality classification, in particular to a mutton quality classification method and system based on multi-source data analysis, and the method comprises the following steps: constructing a distribution curve mapping feature space based on fatty acid data to generate a variety attribution region, constructing a decoupling coefficient characterization shearing force by using a spectral peak difference, and constructing a multi-source data analysis model; and calculating the molar ratio of trace elements to construct a competition vector, solving a steady-state deviation angle to judge the pollution risk, and matching multi-dimensional features to generate a quality classification result. According to the method, the distribution characteristics between the chain length and the unsaturation degree are constructed and mapped to the high-dimensional space to distinguish the variety attribution areas, the shear force intensity is quantified by combining the difference distance between near-infrared characteristic peaks, the metabolic steady-state deviation angle is constructed by using the molar ratio between antagonistic elements, and the pollution risk level is evaluated accordingly. In the classification process, a multi-source heterogeneous data fusion and quantitative feature extraction mechanism is introduced, so that the stable judgment capability of the mutton quality and the judgment precision of the pollution risk are enhanced.

Description

Technical Field

[0001] This invention relates to the field of quality classification technology, and in particular to a method and system for classifying the quality of mutton based on multi-source data analysis. Background Technology

[0002] The field of quality classification technology involves the quantitative assessment and grading of multi-dimensional quality indicators, including physical, chemical, and biological aspects, for agricultural products, food, and industrial products. It encompasses key components such as image analysis, spectral detection, sensor data acquisition, feature extraction, quality assessment standard construction, and classification algorithms. The development of this field relies on precise perception and evaluation mechanisms of sample quality attributes. It typically combines multi-dimensional detection methods with intelligent analysis techniques to model sample attributes and classify quality levels. It is widely applied in various scenarios, including food safety testing, agricultural product quality evaluation, and industrial product quality control, and is characterized by complex data sources, diverse assessment dimensions, and the need for quantitative classification standards.

[0003] Traditional mutton quality classification methods rely on subjective judgment or single-factor measurement of characteristics such as appearance, odor, texture, color, moisture content, and fat distribution through manual experience or a single testing method. These methods often employ manual grading, sensory evaluation, or judgment based on a single data source such as near-infrared spectroscopy, electronic nose, or image recognition. In practice, these methods typically rely on specific types of data collected by a particular testing device; for example, image data is used to determine meat color and fat distribution, near-infrared spectroscopy is used to detect moisture and protein content, and electrochemical sensors are used to detect odor components. Quality is then classified using preset thresholds or simple rules. However, due to the limitations of single data sources in characterization, they cannot comprehensively reflect the multidimensional attributes of mutton quality, thus exhibiting certain deficiencies in classification accuracy and reliability. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose a method and system for classifying mutton quality based on multi-source data analysis.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for classifying mutton quality based on multi-source data analysis, comprising the following steps: S1: Based on the fatty acid data of mutton, divide the orthogonal matrix grid, calculate the concentration quotient, generate the chain-specific unsaturation index, arrange the chain-specific unsaturation index to construct the chain length-unsaturation distribution curve, extract the slope nodes of the chain length-unsaturation distribution curve and map them to a high-dimensional feature space, use support vector machine to divide the high-dimensional feature space, and generate the mutton variety belonging region. S2: Collect near-infrared spectral data of mutton, locate the characteristic peaks of water and protein, calculate the difference between the characteristic peaks and the standard wavelength, construct the difference into a two-dimensional vector and calculate its modulus, construct the water-protein decoupling coefficient, and generate the physical shear force value characterization of muscle using a preset nonlinear segmented mapping rule. S3: Obtain the concentration set of essential trace elements and toxic heavy metals in mutton, calculate and aggregate the molar ratio of antagonistic element pairs, and construct an element competition feature vector. S4: Based on the element competition feature vector, call the standard steady-state feature vector of healthy mutton and calculate the cosine similarity with the element competition feature vector, solve the metabolic steady-state deviation angle, compare the metabolic steady-state deviation angle with the gradient threshold, and generate the pollution risk level; S5: Based on the mutton breed's region, muscle physical shear force value, and contamination risk level, retrieve matching features from the multidimensional quality database to generate a comprehensive mutton quality classification result.

[0006] The present invention improves upon the following: the mutton breed attribution region includes the feature space decision boundary coordinates, the geometric location of the cluster center, and the breed confidence probability density; the muscle physical shear force value characterization includes the muscle fiber fracture strength index, connective tissue hardness parameter, and meat tenderness quantification value; the element competition feature vector includes the ion channel occupancy ratio, metabolic antagonism strength value, and bioavailability competition index; the pollution risk level includes the heavy metal exposure toxicity level, metabolic homeostasis disruption level, and food safety warning category; and the comprehensive mutton quality classification result includes sensory flavor characteristic description, nutritional value grading label, and commercial market grade code.

[0007] The present invention improves upon the following: the mutton breed attribution region includes the feature space decision boundary coordinates, the geometric location of the cluster center, and the breed confidence probability density; the muscle physical shear force value characterization includes the muscle fiber fracture strength index, connective tissue hardness parameter, and meat tenderness quantification value; the element competition feature vector includes the ion channel occupancy ratio, metabolic antagonism strength value, and bioavailability competition index; the pollution risk level includes the heavy metal exposure toxicity level, metabolic homeostasis disruption level, and food safety warning category; and the comprehensive mutton quality classification result includes sensory flavor characteristic description, nutritional value grading label, and commercial market grade code.

[0008] The present invention is improved in that the step of obtaining the muscle physical shear force value characterization quantity is specifically as follows: S211: Collect near-infrared spectral data of mutton, scan and locate the center wavelength of the first-order overtone characteristic peak of the OH bond of water molecules and the center wavelength of the overtone characteristic peak of the NH bond of protein in the full band range, retrieve the reference wavelength of the standard pure water spectral peak and the reference wavelength of the standard undegraded protein peak in the preset database, perform the subtraction difference operation between the measured characteristic peak wavelength and the corresponding reference wavelength for water molecules and protein components respectively, calculate the relative offset value of the two on the spectral wavenumber axis, and generate a set of spectral characteristic peak frequency shift deviations; S212: Based on the set of frequency shift deviations of the spectral characteristic peaks, extract the values ​​of water drift and protein drift as orthogonal coordinate components of the two-dimensional feature space, perform Euclidean distance operation in the vector space, solve the displacement modulus of the coordinate point relative to the origin, characterize the dynamic separation degree of the spectral response of water molecules relative to the spectral response of proteins, and construct the water-protein decoupling coefficient. S213: For the water-protein decoupling coefficient, a preset nonlinear segmented mapping rule is invoked. Based on the magnitude of the water-protein decoupling coefficient, the tenderness conversion interval is determined. The slope parameter and intercept parameter of the interval are retrieved. A linear transformation and weighted summation operation are performed on the water-protein decoupling coefficient to generate a muscle physical shear force value characterization quantity.

[0009] The present invention is improved in that the step of obtaining the element competition feature vector is specifically as follows: S311: Obtain the concentration set of essential trace elements and toxic heavy metals in mutton, call the preset element atomic mass standard parameter library, retrieve the relative atomic mass constants of the corresponding elements of calcium, iron, zinc, lead, cadmium and arsenic, perform the division operation between the mass concentration value and the relative atomic mass constant for each essential trace element and toxic heavy metal, and convert the mass-based concentration index into the molar concentration index based on the number of particles through unit conversion, and establish a dataset of trace element and heavy metal molar concentration. S312: Based on the aforementioned trace element and heavy metal molar concentration dataset, retrieve the preset biological antagonistic competition channel mapping topology, identify and lock essential trace elements and toxic heavy metals that have competitive ion channel occupancy relationships, pair the locked elements to form antagonistic element pairs, extract the essential trace element molar concentration as the numerator and the toxic heavy metal molar concentration as the denominator for each antagonistic element pair, perform point-to-point ratio calculation, quantify the dose suppression intensity between the two, and generate a set of antagonistic element pair molar ratios. S313: For the set of molar ratios of the antagonistic element pairs, a preset feature vector dimension definition template is called. Based on the atomic number or chemical activity sequence of the element pairs, each molar ratio value in the set is mapped and filled into the corresponding dimension coordinate axis position of the multidimensional feature space. The filled numerical sequence is standardized based on the maximum and minimum values ​​to eliminate the influence of the order of magnitude difference between the differentiated element pairs on the vector directionality and construct the element competition feature vector.

[0010] The present invention is improved in that the step of obtaining the pollution risk level is specifically as follows: S411: Based on the element competition feature vector, call the preset healthy mutton standard steady-state feature vector, perform vector point-to-point Euclidean distance operation in multi-dimensional space, retrieve and extract the background ion interference intensity of the detection system, the instrument zero-point reference offset and the effective linear dynamic range of the feature space, and establish a set of vector space geometric parameters. S412: Based on the set of geometric parameters of the vector space, extract the corresponding dimension components of the element competition feature vector and the standard steady-state feature vector of healthy mutton, and combine the background ion interference intensity, the instrument zero-point reference offset, the Euclidean distance between vectors and the effective linear dynamic range of the feature space to calculate and obtain the metabolic steady-state deviation angle. S413: For the metabolic homeostasis deviation angle, retrieve the preset pollution risk assessment grading standard, obtain the gradient threshold sequence corresponding to the risk level, map the deviation angle value to the value range defined by the gradient threshold sequence, determine the severity of heavy metal contamination of the sample based on the position of the value in the range, and generate a pollution risk level.

[0011] The present invention is improved in that the formula for obtaining the metabolic homeostasis deviation angle is specifically as follows: ; in, Represents the angle of deviation from metabolic homeostasis. The eigenvector representing the competing elements Normalized values ​​of the dimensional components, The first eigenvector representing the standard steady-state eigenvector of healthy mutton Normalized values ​​of the dimensional components, This represents the total number of dimensions of the feature vector. This represents the normalized value of the background ion interference intensity of the detection system. The normalized value of the reference offset representing the instrument's zero point. The normalized Euclidean distance between the competitive eigenvectors of representative elements and the standard steady-state eigenvectors of healthy mutton is given. The normalized value representing the effective linear dynamic range of the feature space.

[0012] The present invention is improved in that the steps for obtaining the comprehensive mutton quality classification results are as follows: S511: Based on the mutton breed's region, muscle physical shear force value, and contamination risk level, a preset feature encoding mapping protocol is invoked to convert the region label into a spatial one-hot encoding vector, map the risk level into a discrete ordinal weight factor, perform normalization scaling on the shear force value, and splice and combine each data component after processing according to a preset dimension order to establish a multi-source quality feature query vector. S512: For the multi-source quality feature query vector, retrieve the preset multi-dimensional quality database, traverse the standard quality template data of each type stored in the database, calculate the weighted Euclidean distance between the query vector and the standard template in the multi-dimensional feature space, extract the candidate template set whose distance value is less than the preset matching threshold, and convert the corresponding distance value into a similarity score through inverse proportional transformation to generate a quality feature matching degree matrix. S513: Based on the quality feature matching degree matrix, perform numerical sorting and maximum value retrieval of the matrix column vectors, lock the feature template index with the highest similarity score, retrieve the quality category definition data and attribute description fields associated with the index, map the data into specific text descriptions and grade codes, implement unified judgment on the multi-dimensional attributes of the sample, and generate a comprehensive mutton quality classification result.

[0013] A mutton quality classification system based on multi-source data analysis, wherein the mutton quality classification system based on multi-source data analysis is used to implement the aforementioned mutton quality classification method based on multi-source data analysis, and the system includes: The metabolic phenotype identification module divides the mutton fatty acid data into an orthogonal matrix grid, calculates the concentration quotient, generates a chain-specific unsaturation index, arranges the chain-specific unsaturation index to construct a chain length-unsaturation distribution curve, extracts the slope nodes of the chain length-unsaturation distribution curve and maps them to a high-dimensional feature space, uses a support vector machine to divide the high-dimensional feature space, and generates the mutton breed attribution region. The structural relaxation characterization module collects near-infrared spectral data of mutton, locates the characteristic peaks of water and protein, calculates the difference between the characteristic peaks and the standard wavelength, constructs the difference into a two-dimensional vector and calculates its modulus, constructs the water-protein decoupling coefficient, and generates the muscle physical shear force characterization quantity using a preset nonlinear segmented mapping rule. The element antagonism calculation module obtains the concentration set of essential trace elements and the concentration set of toxic heavy metals in mutton, calculates and aggregates the molar ratio of antagonistic element pairs, and constructs an element competition feature vector. The pollution risk assessment module, based on the element competition feature vector, calls the standard steady-state feature vector of healthy mutton and calculates the cosine similarity with the element competition feature vector, solves the metabolic steady-state deviation angle, compares the metabolic steady-state deviation angle with the gradient threshold, and generates a pollution risk level. The quality grade mapping module, based on the mutton breed's region, muscle physical shear force value, and contamination risk level, retrieves matching features from a multidimensional quality database to generate a comprehensive mutton quality classification result.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the distribution characteristics between chain length and unsaturation degree are constructed and mapped to a high-dimensional space to distinguish the regions to which varieties belong. The difference distance between near-infrared characteristic peaks is combined to quantify shear force intensity. The molar ratio between antagonistic elements is used to construct the metabolic steady-state deviation angle and assess the contamination risk level accordingly. A multi-source heterogeneous data fusion and quantitative feature extraction mechanism is introduced in the classification process, which solves the problem that a single data type cannot fully characterize the multidimensional attributes of mutton. It avoids classification bias caused by subjective judgment or fixed thresholds. By constructing a comprehensive evaluation index, the comprehensiveness, distinguishability and consistency of the classification results are improved, and the ability to stably determine the quality of mutton and the accuracy of the judgment of contamination risk are enhanced. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a flowchart illustrating the process of obtaining the mutton breed classification region according to the present invention; Figure 3 This is a flowchart illustrating the process of obtaining the physical shear force value of muscle according to the present invention; Figure 4 This is a flowchart illustrating the process of obtaining the element competition feature vector in this invention; Figure 5 This is a flowchart for obtaining pollution risk levels in this invention; Figure 6 This is a flowchart illustrating the process of obtaining comprehensive mutton quality classification results according to the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0018] Please see Figure 1 The present invention provides a technical solution, a method for classifying mutton quality based on multi-source data analysis, comprising the following steps; S1: Obtain fatty acid detection data of mutton, divide the orthogonal matrix grid according to carbon chain length and double bond number, calculate the weighted sum of unsaturated isomer concentrations under the target chain length and the quotient of saturated fatty acid concentration, generate chain-specific unsaturation index, arrange the chain-specific unsaturation index in order of carbon atom number, construct chain length-unsaturation distribution curve, extract the three nodes with the largest absolute value of change slope by differentiating the chain length-unsaturation distribution curve, map the node slope value and carbon chain position to high-dimensional feature space, call support vector machine to perform hyperplane partitioning on high-dimensional feature space, and generate mutton variety classification region; S2: Collect near-infrared spectral data of mutton, locate the first-order overtone characteristic peak of the OH bond of water molecules and the overtone characteristic peak of the NH bond of protein, calculate the difference between the center wavelength of the first-order overtone characteristic peak and the standard pure water spectral peak and the difference between the center wavelength of the overtone characteristic peak and the standard undegraded protein peak, calculate the Euclidean distance between the two differences, construct the water-protein decoupling coefficient, input the water-protein decoupling coefficient into the nonlinear piecewise mapping rule, and generate the muscle physical shear force value characterization quantity; S3: Obtain the concentration set of essential trace elements and the concentration set of toxic heavy metals in mutton. For antagonistic element pairs consisting of one essential trace element and one toxic heavy metal, perform a molar concentration division operation, aggregate the molar ratio of all antagonistic element pairs, and construct an element competition feature vector. S4: Based on the element competition feature vector, call the preset healthy mutton standard steady-state feature vector, perform vector space cosine similarity calculation, solve the metabolic steady-state deviation angle between the element competition feature vector and the standard steady-state feature vector, compare the metabolic steady-state deviation angle with the preset gradient threshold, and generate the pollution risk level. S5: Based on the mutton breed classification region, muscle physical shear force value, and contamination risk level, a multidimensional quality database is searched for feature matching to generate a comprehensive mutton quality classification result. The classification of mutton breeds includes the coordinates of the decision boundary in the feature space, the geometric location of the cluster center, and the probability density of breed confidence. The physical shear force values ​​of muscle include the muscle fiber fracture strength index, connective tissue hardness parameters, and quantitative values ​​of meat tenderness. The element competition feature vector includes the ion channel occupancy ratio, metabolic antagonism intensity value, and bioavailability competition index. The pollution risk level includes the heavy metal exposure toxicity level, metabolic homeostasis disruption level, and food safety warning category. The comprehensive mutton quality classification results include sensory flavor characteristics descriptions, nutritional value grading labels, and commercial market grade codes.

[0019] Please see Figure 2 The specific steps for determining the territorial classification of mutton breeds are as follows: S111: Obtain mutton fatty acid detection data, call the preset orthogonal matrix of carbon chain length and double bond number, map the mutton fatty acid detection data to the grid cells of the orthogonal matrix, perform concentration value aggregation summation for unsaturated isomers in each grid cell, perform division operation between the aggregation summation value under the target chain length and the saturated fatty acid concentration of the corresponding chain length, and generate chain-specific unsaturation index. The chromatographic peak areas and retention times of thirty different fatty acids were determined using gas chromatography-mass spectrometry (GC-MS). A preset orthogonal matrix of carbon chain length and double bond number was used, which is divided into rows based on the number of carbon atoms (e.g., C14 to C24) and columns based on the number of double bonds (e.g., 0 to 6). Each detected fatty acid is mapped to a corresponding grid cell based on its chemical structure characteristics. For example, the conjugated linoleic acid isomer is mapped to a grid cell with 18 rows and 2 columns of double bonds. Concentration values ​​are aggregated and summed for all unsaturated isomers within each grid cell. The grid cell is set to contain [various components] within the C18 chain length. -CLA、 -CLA and - Linoleic acid has three isomers, with concentrations of respectively , and The aggregate summation value is then... Subsequently, the concentration of saturated fatty acids with the corresponding C18 chain length (i.e., stearic acid C18:0) was extracted and set as [value missing]. Perform division operation Repeat the above calculation for all carbon chain lengths to generate the chain-specific unsaturation index.

[0020] S112: Based on the chain-specific unsaturation index, the chain length-unsaturation distribution curve is generated by arranging the discrete data points according to the increasing sequence of carbon atoms, connecting the discrete data points, performing a first-order discrete difference operation on the chain length-unsaturation distribution curve, obtaining the slope of change between adjacent nodes, extracting key nodes whose absolute value of the slope of change is greater than a preset threshold, obtaining the slope value of the node and the corresponding carbon chain position parameters, and generating the topological mapping coordinates of the feature nodes. Arranged according to the increasing sequence of carbon atoms from C14 to C24, the discrete data points are connected to generate a chain length-unsaturation distribution curve, as shown in Table 1 for a sample data distribution. A first-order discrete difference operation is performed on the chain length-unsaturation distribution curve, and the calculation formula is as follows: For example, if the exponent is 0.3 at C16 and increases to 0.35 at C17, then the slope of change in this segment (after considering step size normalization) is: The entire curve is traversed to obtain the slope of change between adjacent nodes. Key nodes with a slope absolute value greater than a preset threshold are extracted and set as C18 (slope 0.45), C16 (slope 0.22), and C20 (slope -0.15). The slope values ​​of these three nodes and the corresponding carbon chain position parameters (such as 18, 16, 20) are obtained and combined into a six-dimensional feature vector to generate the topological mapping coordinates of the feature nodes.

[0021] Table 1. Example of chain length-unsaturation distribution data; S113: Call the support vector machine model, input the topological mapping coordinates of the feature nodes into the support vector machine model, map the coordinates to the high-dimensional feature space based on the preset kernel function, calculate the geometric margin between each sample point in the high-dimensional feature space and the optimal classification hyperplane, determine the region based on the sign attribute and numerical range of the geometric margin, and generate the region to which the mutton variety belongs.

[0022] Based on a pre-defined radial basis kernel function, the six-dimensional coordinates are mapped to a high-dimensional feature space. The geometric margin between each sample point in the high-dimensional feature space and the optimal classification hyperplane is calculated. For example, for the input vector... The decision function is If the calculated decision value is And the preset area A determination range is The range of region B is as follows: Based on the sign attribute (positive) and numerical range (greater than 1.0) of the geometric interval, the region is determined to be within the feature cluster corresponding to the Xinjiang sheep production area, thus generating the mutton breed affiliation region.

[0023] Please see Figure 3 The specific steps for obtaining the physical shear force value of muscle are as follows: S211: Collect near-infrared spectral data of mutton, scan and locate the center wavelength of the first-order overtone characteristic peak of the OH bond of water molecules and the center wavelength of the overtone characteristic peak of the NH bond of protein in the full band range, retrieve the reference wavelength of the standard pure water spectral peak and the reference wavelength of the standard undegraded protein peak in the preset database, perform the subtraction difference operation between the measured characteristic peak wavelength and the corresponding reference wavelength for water molecules and protein components respectively, calculate the relative offset value of the two on the spectral wavenumber axis, and generate a set of spectral characteristic peak frequency shift deviations; exist to Scanning across the entire wavelength range, the center wavelength of the characteristic peak in the first overtone region corresponding to the OH bond of the water molecule was located using the second derivative method (measured value). The center wavelength of the overtone characteristic peak of the NH bond in the protein (measured value) and the center wavelength of the overtone characteristic peak of the protein NH bond. ), retrieve the reference wavelength of the standard pure water spectral peak from the preset database (set to ), ) and the reference wavelength of the standard undegraded protein peak (set to) The reference wavelength is set based on The statistical mean of the standard spectral response of pure substances in the environment was used to calculate the frequency shift of water molecules by performing subtraction difference operations between the measured characteristic peak wavelength and the corresponding reference wavelength for both water molecules and protein components. Protein frequency shift The relative offset values ​​of the two on the spectral wavenumber axis are calculated to generate a set of spectral characteristic peak frequency shift deviations.

[0024] S212: Based on the set of frequency shift deviations of spectral characteristic peaks, the values ​​of water drift and protein drift are extracted as orthogonal coordinate components in the two-dimensional feature space. Euclidean distance operation is performed in the vector space to solve the displacement modulus of the coordinate point relative to the origin. This characterizes the dynamic separation degree of the spectral response of water molecules relative to the spectral response of proteins, quantifies the change in chemical bond constant caused by water escaping from the protein network, and constructs the water-protein decoupling coefficient. Extract water drift value ( ) and protein drift value ( (as orthogonal coordinate components of a two-dimensional feature space) Perform Euclidean distance calculation in the vector space, the formula is as follows: The numerical solution calculates the displacement modulus of the coordinate point relative to the origin, characterizes the dynamic separation degree of the spectral response of water molecules relative to the spectral response of proteins, quantifies the change in chemical bond force constant caused by water escaping from the protein network, and constructs the water-protein decoupling coefficient.

[0025] S213: For the water-protein decoupling coefficient, the preset nonlinear segmented mapping rule is invoked. Based on the magnitude of the water-protein decoupling coefficient, the tenderness conversion interval is determined. The slope parameter and intercept parameter of the interval are retrieved. Linear transformation and weighted summation are performed on the water-protein decoupling coefficient. The microscopic spectral decoupling characteristics are inverted and mapped into macroscopic mechanical property values ​​to characterize the relaxation degree of muscle fiber structure and generate a muscle physical shear force value characterization quantity.

[0026] The rule is set when the coefficient is less than It belongs to the high tenderness range, greater than or equal to and less than If the sample falls within the medium tenderness range, the medium tenderness transition range is determined based on the magnitude of the water-protein decoupling coefficient, and the slope parameter of that range is retrieved. and intercept parameter The parameters were obtained through regression fitting of destructive shearing experimental data from 500 historical samples. A linear transformation and weighted summation were performed on the water-protein decoupling coefficients. The calculation process is as follows: The microscopic spectral decoupling features are inverted and mapped into macroscopic mechanical property values, characterizing the degree of relaxation of muscle fiber structure and generating a physical shear force value characterization quantity for muscles.

[0027] Please see Figure 4 The specific steps for obtaining the element competition feature vector are as follows: S311: Obtain the concentration set of essential trace elements and toxic heavy metals in mutton, call the preset element atomic mass standard parameter library, retrieve the relative atomic mass constants of the corresponding elements of calcium, iron, zinc, lead, cadmium and arsenic, perform the division operation between the mass concentration value and the relative atomic mass constant for each essential trace element and toxic heavy metal, and convert the mass-based concentration index into the molar concentration index based on the number of particles through unit conversion, and establish a dataset of trace element and heavy metal molar concentration. For example, the zinc (Zn) concentration was detected as follows: The cadmium (Cd) concentration was The system calls a pre-defined library of standard atomic mass parameters for elements, retrieving the relative atomic mass constants for calcium, iron, zinc, lead, cadmium, and arsenic. The values ​​are Zn = 65.38, Cd = 112.41, and the system performs a division operation between the mass concentration value and the relative atomic mass constant for each essential trace element and toxic heavy metal. The approximate molar concentration of zinc is... The approximate molar concentration of cadmium is By converting the mass-based concentration index into the particle-based molar concentration index through unit conversion, a dataset of molar concentrations of trace elements and heavy metals is established.

[0028] S312: Based on the dataset of trace element and heavy metal molar concentrations, retrieve the preset biological antagonistic competitive channel mapping topology, identify and lock essential trace elements and toxic heavy metals that have competitive ion channel occupancy relationships, pair the locked elements to form antagonistic element pairs, extract the essential trace element molar concentration as the numerator and the toxic heavy metal molar concentration as the denominator for each antagonistic element pair, perform point-to-point ratio calculation, quantify the dose suppression intensity between the two, and generate a set of antagonistic element pair molar ratios. Biological antagonistic competitive channel mapping topology defines element combinations that share divalent metal ion transporters (DMT1), such as "zinc-cadmium", "calcium-lead", and "iron-arsenic". It identifies and locks essential trace elements (such as zinc) and toxic heavy metals (such as cadmium) that have competitive ion channel occupancy relationships. The locked elements are paired to form antagonistic element pairs. For each antagonistic element pair, the molar concentration of the essential trace element is extracted. ) as the numerator and the molar concentration of toxic heavy metals ( As the denominator, a point-to-point ratio calculation is performed, i.e. This value quantifies the dose suppression intensity between the two, generating a set of antagonistic element pair molar ratios.

[0029] S313: For the set of molar ratios of antagonistic element pairs, a preset feature vector dimension definition template is called. Based on the atomic number or chemical activity sequence of the element pairs, each molar ratio value in the set is mapped and filled into the corresponding dimension coordinate axis position of the multidimensional feature space. The filled numerical sequence is standardized based on the maximum and minimum values ​​to eliminate the influence of the order of magnitude difference between the differentiated element pairs on the vector directionality and construct the element competition feature vector.

[0030] Given a set containing [Zn / Cd: 1563.6, Ca / Pb: 850.2, Fe / As: 1200.5], a preset feature vector dimension definition template is used. Based on the atomic number sequence of the element pairs, each molar ratio value in the set is sequentially mapped and filled into the corresponding coordinate axis position of the multidimensional feature space. The filled numerical sequence is then normalized based on the maximum and minimum values. The historical maximum value for the Zn / Cd dimension is set to 2000, and the minimum value to 100. The normalized value is then... Similarly, other dimensions are processed to eliminate the influence of the order-of-magnitude difference between differential element pairs on the vector directionality, and to construct element-competitive feature vectors.

[0031] Please see Figure 5 The specific steps for obtaining the pollution risk level are as follows: S411: Based on the element competition feature vector, call the preset healthy mutton standard steady-state feature vector, perform vector point-to-point Euclidean distance operation in multi-dimensional space, retrieve and extract the background ion interference intensity of the detection system, the instrument zero-point reference offset and the effective linear dynamic range of the feature space, and establish a set of vector space geometric parameters. For example, for a two-dimensional simplified vector Call the preset healthy mutton standard steady-state feature vector (e.g. ), perform point-to-point Euclidean distance calculations in multidimensional space, and retrieve and extract the background ion interference intensity of the detection system ( ), Instrument zero-point reference offset ( ) and the effective linear dynamic range of the feature space ( ), and establish the set of geometric parameters for the vector space.

[0032] S412: Based on the geometric parameter set of the vector space, extract the corresponding dimensional components of the element competition feature vector and the standard steady-state feature vector of healthy mutton. Combined with the background ion interference intensity, instrument zero-point reference offset, Euclidean distance between vectors and the effective linear dynamic range of the feature space, the formula is used: ; The metabolic homeostasis deviation angle is obtained through calculation; in, Represents the angle of deviation from metabolic homeostasis. The eigenvector representing the competing elements The normalized values ​​of the dimensional components are obtained by extracting the molar ratios of element pairs and normalizing them based on the extreme values ​​of the dataset. The first eigenvector representing the standard steady-state eigenvector of healthy mutton The normalized values ​​of the dimensional components are obtained by retrieving and normalizing them from a pre-defined database of standard healthy samples. The total dimension of the feature vector is determined by the count of antagonistic element pairs. The normalized value representing the background ion interference intensity of the detection system is obtained by measuring and normalizing the response signal of the blank matrix solution. The normalized value of the reference offset representing the instrument's zero point is obtained by acquiring baseline drift data through the zero-point calibration procedure and then normalizing it. The normalized Euclidean distance between the competitive eigenvectors of the representative elements and the standard steady-state eigenvectors of healthy mutton is obtained by calculating the square root of the sum of the squares of the differences in the coordinates of the two vectors. The effective linear dynamic range normalized value representing the feature space is obtained by calculating the upper and lower limits of the detector's linear response interval and normalizing it. The gradient threshold sequence is obtained by constructing quantile nodes based on the statistical distribution characteristics and toxicological tolerance limits of historical mutton sample contamination data. formula: ; By introducing cosine similarity to calculate the directional consistency of two vectors, and combining it with instrument noise ( ) and zero-point drift ( Signal correction is performed using Euclidean distance ( ) ) and dynamic range ( The ratio of θ to θ is used as a penalty term, and spatial distance weighting is applied to the pure angular deviation to more accurately quantify the degree of metabolic deviation. Represents the angle of deviation from metabolic homeostasis; The eigenvector representing the competing elements The normalized values ​​of the dimensional components are obtained as calculated above. ; The first eigenvector representing the standard steady-state eigenvector of healthy mutton The normalized value of the dimensional component, taking values ​​of ; This represents the total number of dimensions of the feature vectors, which is 2 in this case. The normalized value representing the background ion interference intensity of the detection system is obtained by measuring the baseline response intensity of the blank solvent and is set as follows. ; The normalized value of the reference offset representing the instrument's zero point is obtained through the power-on calibration procedure and set as follows. ; The normalized Euclidean distance between two vectors is calculated as follows: ; The normalized value representing the effective linear dynamic range of the feature space is set to . .

[0033] The specific calculation process is as follows: Calculate the dot product term: ; Calculate the modulus term: ; ; Calculate the denominator: ; Calculate the numerator: ; Calculate the cosine value and angle: ; (about ); Calculate the penalty coefficient: ; Final result: ; This result indicates that the elemental metabolic pattern of the sample differs from the health standard by approximately [missing information]. The overall deviation in radians reflects the degree of disturbance of the trace element metabolic network by heavy metals, and the metabolic homeostasis deviation angle is calculated.

[0034] S413: For the metabolic homeostasis deviation angle, retrieve the preset pollution risk assessment grading standard, obtain the gradient threshold sequence corresponding to the risk level, map the deviation angle value to the numerical range defined by the gradient threshold sequence, determine the severity of heavy metal contamination of the sample based on the position of the value in the range, and generate the pollution risk level.

[0035] Regarding the deviation angle of metabolic homeostasis ( The gradient threshold sequence includes a first-level risk threshold. Level 2 risk threshold Level 3 risk threshold The threshold was set as follows: based on historical data from 1000 mutton samples, the deviation angle distribution was statistically analyzed, and the 25th, 50th, and 75th percentile values ​​were selected. This data was then combined with the elemental concentrations corresponding to the tolerable daily intake (TDI) in toxicology for inversion calculation. For example... Corresponding to the deviation state when the cadmium content in the sample reaches 50% of the national standard limit, the deviation angle value is... Mapped to the numerical range defined by the gradient threshold sequence, because The value is determined to fall within the range between Level 1 and Level 2 risk, i.e., the "slight pollution risk" level. The severity of heavy metal contamination of the sample is determined based on the position of the value within the range, and a pollution risk level is generated.

[0036] Please see Figure 6 The specific steps for obtaining the comprehensive mutton quality classification results are as follows: S511: Based on the mutton breed's region, muscle physical shear force value, and pollution risk level, a preset feature coding mapping protocol is invoked to convert the region label into a spatial one-hot coding vector, map the risk level into a discrete ordinal weight factor, perform normalization scaling on the shear force value, and splice and combine each data component after processing according to a preset dimension order to establish a multi-source quality feature query vector. Based on the mutton breed's geographical origin (e.g., "Xinjiang"), the muscle physical shear force value characterization ( The system uses the pollution risk level ("mild risk") and calls a preset feature encoding mapping protocol, for example, mapping "Xinjiang" to... "Ujumqin" is mapped to This maps risk levels to discrete ordinal weighting factors; for example, "mild risk" is mapped to a numerical value. "Severe risk" is mapped to The shear force characterization was normalized and scaled, and the maximum shear force was set to... The normalized value is Each processed data component is concatenated and combined according to a preset dimensional order to construct a vector. Establish a multi-source quality feature query vector.

[0037] S512: For the multi-source quality feature query vector, retrieve the preset multi-dimensional quality database, traverse the standard quality template data of each type stored in the database, calculate the weighted Euclidean distance between the query vector and the standard template in the multi-dimensional feature space, extract the candidate template set whose distance value is less than the preset matching threshold, and convert the corresponding distance value into a similarity score through inverse proportional transformation to generate a quality feature matching degree matrix. The preset matching threshold is obtained by selecting historically confirmed mutton quality standard samples to construct a training dataset. For different samples under the same quality category, the intra-class weighted Euclidean distance set between their feature vectors is calculated. The distance set is fitted with a normal distribution to obtain the mean and standard deviation parameters. The upper limit critical value of the distribution is calculated based on the preset statistical confidence level, and the statistical upper limit critical value is used as the matching threshold for judging feature similarity. For example, selecting 50 samples under the category of "Premium Xinjiang Lamb", calculating the intra-class weighted Euclidean distance set between their feature vectors, the mean of this distance set is calculated to be... The standard deviation is Based on the preset 95% statistical confidence level (corresponding to ), calculate the upper critical value of the distribution. And set the statistical upper limit critical value As a matching threshold for determining feature similarity, if the distance between a query vector and the template is... less than If the match is valid, then the match is considered valid.

[0038] S513: Based on the quality feature matching degree matrix, perform numerical sorting and maximum value retrieval of the matrix column vectors, lock the feature template index with the highest similarity score, retrieve the quality category definition data and attribute description fields associated with the index, map the data into specific text descriptions and grade codes, implement unified judgment on the multi-dimensional attributes of the sample, and generate comprehensive mutton quality classification results.

[0039] The matrix contains the similarity scores of each candidate template, for example, {"Premium A": 0.92, "Level B": 0.85, "Level C": 0.40}. The matrix column vectors are sorted numerically and the maximum value is retrieved to lock the feature template index with the highest similarity score (corresponding to "Premium A"). The quality category definition data and attribute description fields associated with the index are retrieved, such as "tender meat, no pollution risk, typical Xinjiang flavor". The data is mapped to specific text descriptions and grade codes (e.g., "Code: S-Level-1"), and a unified judgment of the multi-dimensional attributes of the samples is implemented to generate a comprehensive mutton quality classification result.

[0040] A mutton quality classification system based on multi-source data analysis is provided. This system implements the aforementioned mutton quality classification method based on multi-source data analysis. The system includes: The metabolic phenotype identification module divides the mutton fatty acid data into an orthogonal matrix grid, calculates the concentration quotient, generates the chain-specific unsaturation index, arranges the chain-specific unsaturation index to construct the chain length-unsaturation distribution curve, extracts the slope nodes of the chain length-unsaturation distribution curve and maps them to a high-dimensional feature space, uses a support vector machine to divide the high-dimensional feature space, and generates the mutton breed classification region. The structural relaxation characterization module collects near-infrared spectral data of mutton, locates the characteristic peaks of water and protein, calculates the difference between the characteristic peaks and the standard wavelength, constructs the difference into a two-dimensional vector and calculates its modulus, constructs the water-protein decoupling coefficient, inputs the water-protein decoupling coefficient into the nonlinear piecewise mapping rule, and generates a characterization quantity of muscle physical shear force. The element antagonism calculation module obtains the concentration set of essential trace elements and the concentration set of toxic heavy metals in mutton, calculates and aggregates the molar ratio of antagonistic element pairs, and constructs an element competition feature vector. The pollution risk assessment module, based on the element competition feature vector, calls the standard steady-state feature vector of healthy mutton and calculates the cosine similarity with the element competition feature vector, solves the metabolic steady-state deviation angle, compares the metabolic steady-state deviation angle with the gradient threshold, and generates the pollution risk level. The quality grade mapping module, based on the mutton breed's region, muscle physical shear force value, and contamination risk level, retrieves matching features from a multidimensional quality database to generate a comprehensive mutton quality classification result.

[0041] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for classifying mutton quality based on multi-source data analysis, characterized in that, Includes the following steps: S1: Based on the fatty acid data of mutton, divide the orthogonal matrix grid, calculate the concentration quotient, generate the chain-specific unsaturation index, arrange the chain-specific unsaturation index to construct the chain length-unsaturation distribution curve, extract the slope nodes of the chain length-unsaturation distribution curve and map them to a high-dimensional feature space, use support vector machine to divide the high-dimensional feature space, and generate the mutton variety belonging region. S2: Collect near-infrared spectral data of mutton, locate the characteristic peaks of water and protein, calculate the difference between the characteristic peaks and the standard wavelength, construct the difference into a two-dimensional vector and calculate its modulus, construct the water-protein decoupling coefficient, and generate the physical shear force value characterization of muscle using a preset nonlinear segmented mapping rule. S3: Obtain the concentration set of essential trace elements and toxic heavy metals in mutton, calculate and aggregate the molar ratio of antagonistic element pairs, and construct an element competition feature vector. S4: Based on the element competition feature vector, call the standard steady-state feature vector of healthy mutton and calculate the cosine similarity with the element competition feature vector, solve the metabolic steady-state deviation angle, compare the metabolic steady-state deviation angle with the gradient threshold, and generate the pollution risk level; S5: Based on the mutton breed's region, muscle physical shear force value, and contamination risk level, retrieve matching features from the multidimensional quality database to generate a comprehensive mutton quality classification result.

2. The method for classifying mutton quality based on multi-source data analysis according to claim 1, characterized in that, The mutton breed attribution region includes the feature space decision boundary coordinates, the geometric location of the cluster center, and the breed confidence probability density. The muscle physical shear force value characterization includes the muscle fiber fracture strength index, connective tissue hardness parameter, and meat tenderness quantification value. The element competition feature vector includes the ion channel occupancy ratio, metabolic antagonism strength value, and bioavailability competition index. The pollution risk level includes the heavy metal exposure toxicity level, metabolic homeostasis disruption level, and food safety warning category. The comprehensive mutton quality classification result includes sensory flavor characteristic description, nutritional value grading label, and commercial market grade code.

3. The method for classifying mutton quality based on multi-source data analysis according to claim 2, characterized in that, The specific steps for determining the mutton breed's region are as follows: S111: Obtain mutton fatty acid detection data, call the preset orthogonal matrix of carbon chain length and double bond number, map the mutton fatty acid detection data to the grid cells of the orthogonal matrix, perform concentration value aggregation summation for unsaturated isomers in each grid cell, perform division operation between the aggregation summation value under the target chain length and the saturated fatty acid concentration of the corresponding chain length, and generate chain-specific unsaturation index. S112: Based on the chain-specific unsaturation index, the chain length-unsaturation distribution curve is generated by arranging the discrete data points according to the increasing sequence of carbon atoms, and performing a first-order discrete difference operation on the chain length-unsaturation distribution curve to obtain the slope of change between adjacent nodes. Key nodes with a slope absolute value greater than a preset threshold are extracted, and the slope value of the node and the corresponding carbon chain position parameters are obtained to generate the topological mapping coordinates of the feature node. S113: Call the support vector machine model, input the topological mapping coordinates of the feature nodes into the support vector machine model, map the coordinates to the high-dimensional feature space based on the preset kernel function, calculate the geometric interval between each sample point in the high-dimensional feature space and the optimal classification hyperplane, determine the region based on the sign attribute and numerical range of the geometric interval, and generate the mutton variety belonging region.

4. The method for classifying mutton quality based on multi-source data analysis according to claim 3, characterized in that, The specific steps for obtaining the muscle physical shear force value characterization are as follows: S211: Collect near-infrared spectral data of mutton, scan and locate the center wavelength of the first-order overtone characteristic peak of the OH bond of water molecules and the center wavelength of the overtone characteristic peak of the NH bond of protein in the full band range, retrieve the reference wavelength of the standard pure water spectral peak and the reference wavelength of the standard undegraded protein peak in the preset database, perform the subtraction difference operation between the measured characteristic peak wavelength and the corresponding reference wavelength for water molecules and protein components respectively, calculate the relative offset value of the two on the spectral wavenumber axis, and generate a set of spectral characteristic peak frequency shift deviations; S212: Based on the set of frequency shift deviations of the spectral characteristic peaks, extract the values ​​of water drift and protein drift as orthogonal coordinate components of the two-dimensional feature space, perform Euclidean distance operation in the vector space, solve the displacement modulus of the coordinate point relative to the origin, characterize the dynamic separation degree of the spectral response of water molecules relative to the spectral response of proteins, and construct the water-protein decoupling coefficient. S213: For the water-protein decoupling coefficient, a preset nonlinear segmented mapping rule is invoked. Based on the magnitude of the water-protein decoupling coefficient, the tenderness conversion interval is determined. The slope parameter and intercept parameter of the interval are retrieved. A linear transformation and weighted summation operation are performed on the water-protein decoupling coefficient to generate a muscle physical shear force value characterization quantity.

5. The method for classifying mutton quality based on multi-source data analysis according to claim 4, characterized in that, The specific steps for obtaining the element competition feature vector are as follows: S311: Obtain the concentration set of essential trace elements and toxic heavy metals in mutton, call the preset element atomic mass standard parameter library, retrieve the relative atomic mass constants of the corresponding elements of calcium, iron, zinc, lead, cadmium and arsenic, perform the division operation between the mass concentration value and the relative atomic mass constant for each essential trace element and toxic heavy metal, and convert the mass-based concentration index into the molar concentration index based on the number of particles through unit conversion, and establish a dataset of trace element and heavy metal molar concentration. S312: Based on the aforementioned trace element and heavy metal molar concentration dataset, retrieve the preset biological antagonistic competition channel mapping topology, identify and lock essential trace elements and toxic heavy metals that have competitive ion channel occupancy relationships, pair the locked elements to form antagonistic element pairs, extract the essential trace element molar concentration as the numerator and the toxic heavy metal molar concentration as the denominator for each antagonistic element pair, perform point-to-point ratio calculation, quantify the dose suppression intensity between the two, and generate a set of antagonistic element pair molar ratios. S313: For the set of molar ratios of the antagonistic element pairs, a preset feature vector dimension definition template is called. Based on the atomic number or chemical activity sequence of the element pairs, each molar ratio value in the set is mapped and filled into the corresponding dimension coordinate axis position of the multidimensional feature space. The filled numerical sequence is standardized based on the maximum and minimum values ​​to eliminate the influence of the order of magnitude difference between the differentiated element pairs on the vector directionality and construct the element competition feature vector.

6. The method for classifying mutton quality based on multi-source data analysis according to claim 5, characterized in that, The specific steps for obtaining the pollution risk level are as follows: S411: Based on the element competition feature vector, call the preset healthy mutton standard steady-state feature vector, perform vector point-to-point Euclidean distance operation in multi-dimensional space, retrieve and extract the background ion interference intensity of the detection system, the instrument zero-point reference offset and the effective linear dynamic range of the feature space, and establish a set of vector space geometric parameters. S412: Based on the set of geometric parameters of the vector space, extract the corresponding dimension components of the element competition feature vector and the standard steady-state feature vector of healthy mutton, and combine the background ion interference intensity, the instrument zero-point reference offset, the Euclidean distance between vectors and the effective linear dynamic range of the feature space to calculate and obtain the metabolic steady-state deviation angle. S413: For the metabolic homeostasis deviation angle, retrieve the preset pollution risk assessment grading standard, obtain the gradient threshold sequence corresponding to the risk level, map the deviation angle value to the value range defined by the gradient threshold sequence, determine the severity of heavy metal contamination of the sample based on the position of the value in the range, and generate a pollution risk level.

7. The method for classifying mutton quality based on multi-source data analysis according to claim 6, characterized in that, The specific formula for obtaining the metabolic homeostasis deviation angle is as follows: ； in, Represents the angle of deviation from metabolic homeostasis. The eigenvector representing the competing elements Normalized values ​​of the dimensional components, The first eigenvector representing the standard steady-state eigenvector of healthy mutton Normalized values ​​of the dimensional components, This represents the total number of dimensions of the feature vector. This represents the normalized value of the background ion interference intensity of the detection system. The normalized value of the reference offset representing the instrument's zero point. The normalized Euclidean distance between the competitive eigenvectors of representative elements and the standard steady-state eigenvectors of healthy mutton is used. The normalized value representing the effective linear dynamic range of the feature space.

8. The method for classifying mutton quality based on multi-source data analysis according to claim 7, characterized in that, The specific steps for obtaining the comprehensive mutton quality classification results are as follows: S511: Based on the mutton breed's region, muscle physical shear force value, and contamination risk level, a preset feature encoding mapping protocol is invoked to convert the region label into a spatial one-hot encoding vector, map the risk level into a discrete ordinal weight factor, perform normalization scaling on the shear force value, and splice and combine each data component after processing according to a preset dimension order to establish a multi-source quality feature query vector. S512: For the multi-source quality feature query vector, retrieve the preset multi-dimensional quality database, traverse the standard quality template data of each type stored in the database, calculate the weighted Euclidean distance between the query vector and the standard template in the multi-dimensional feature space, extract the candidate template set whose distance value is less than the preset matching threshold, and convert the corresponding distance value into a similarity score through inverse proportional transformation to generate a quality feature matching degree matrix. S513: Based on the quality feature matching degree matrix, perform numerical sorting and maximum value retrieval of the matrix column vectors, lock the feature template index with the highest similarity score, retrieve the quality category definition data and attribute description fields associated with the index, map the data into specific text descriptions and grade codes, implement unified judgment on the multi-dimensional attributes of the sample, and generate a comprehensive mutton quality classification result.

9. A mutton quality classification system based on multi-source data analysis, characterized in that, The system is used to implement the mutton quality classification method based on multi-source data analysis as described in any one of claims 1-8, and the system comprises: The metabolic phenotype identification module divides the mutton fatty acid data into an orthogonal matrix grid, calculates the concentration quotient, generates a chain-specific unsaturation index, arranges the chain-specific unsaturation index to construct a chain length-unsaturation distribution curve, extracts the slope nodes of the chain length-unsaturation distribution curve and maps them to a high-dimensional feature space, uses a support vector machine to divide the high-dimensional feature space, and generates the mutton breed attribution region. The structural relaxation characterization module collects near-infrared spectral data of mutton, locates the characteristic peaks of water and protein, calculates the difference between the characteristic peaks and the standard wavelength, constructs the difference into a two-dimensional vector and calculates its modulus, constructs the water-protein decoupling coefficient, and generates the muscle physical shear force characterization quantity using a preset nonlinear segmented mapping rule. The element antagonism calculation module obtains the concentration set of essential trace elements and the concentration set of toxic heavy metals in mutton, calculates and aggregates the molar ratio of antagonistic element pairs, and constructs an element competition feature vector. The pollution risk assessment module, based on the element competition feature vector, calls the standard steady-state feature vector of healthy mutton and calculates the cosine similarity with the element competition feature vector, solves the metabolic steady-state deviation angle, compares the metabolic steady-state deviation angle with the gradient threshold, and generates a pollution risk level. The quality grade mapping module, based on the mutton breed's region, muscle physical shear force value, and contamination risk level, retrieves matching features from a multidimensional quality database to generate a comprehensive mutton quality classification result.

Images