CN-121997153-A - Mutton quality classification method and system based on multi-source data analysis

CN121997153ACN 121997153 ACN121997153 ACN 121997153ACN-121997153-A

Abstract

The invention relates to the technical field of quality classification, in particular to a mutton quality classification method and a mutton quality classification system based on multi-source data analysis, which comprise the following steps of constructing a distribution curve mapping characteristic space based on fatty acid data to generate a variety attribution region, and constructing decoupling coefficients to represent shearing force by utilizing spectrum peak differences, calculating the mole ratio of trace elements to construct competition vectors, solving steady-state departure angles to judge pollution risks, and matching multidimensional features to generate quality classification results. According to the invention, the distribution characteristics between chain length and unsaturation degree are constructed and mapped to a high-dimensional space to realize the distinction of variety attribution areas, the difference distance between near infrared characteristic peaks is combined to quantify the shear force intensity, the metabolic steady-state deviation angle is constructed by utilizing the mole ratio between antagonistic elements, the pollution risk level is evaluated according to the metabolic steady-state deviation angle, a multi-source heterogeneous data fusion and quantification characteristic extraction mechanism is introduced in the classification process, and the stable determination capability of mutton quality and the determination precision of pollution risk are enhanced.

Inventors

GONG PING
WU LAN
HU XIN
Chai ting
WEI PEILING
GAO WEIMING
ZHANG RONGYIN
XU YANLI
WANG XIAOTAO
Guan Mingxuan
ZHONG LIWEI

Assignees

新疆维吾尔自治区畜牧科学院

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (9)

1. The mutton quality classification method based on multi-source data analysis is characterized by comprising the following steps of: s1, dividing an orthogonal matrix grid based on mutton fatty acid data, calculating a concentration quotient, generating a chain-specific unsaturated degree index, arranging the chain-specific unsaturated degree index to construct a chain length-unsaturated degree distribution curve, extracting a slope node of the chain length-unsaturated degree distribution curve, mapping the slope node to a high-dimensional characteristic space, and dividing the high-dimensional characteristic space by using a support vector machine to generate a mutton variety attribution region; S2, collecting mutton near infrared spectrum data, positioning characteristic peaks of water and protein, calculating a difference value between the characteristic peaks and standard wavelength, constructing a two-dimensional vector by the difference value, calculating the module length of the two-dimensional vector, constructing a water-protein decoupling coefficient, and generating a physical shear force value characterization quantity of muscle by utilizing a preset nonlinear piecewise mapping rule; S3, acquiring concentration sets of trace elements and toxic heavy metal necessary for mutton, calculating the mole ratio of antagonistic elements to each other, and polymerizing to construct element competition characteristic vectors; s4, based on the element competition feature vector, calling a healthy mutton standard steady-state feature vector, calculating cosine similarity with the element competition feature vector, solving a metabolic steady-state deviation angle, and comparing the metabolic steady-state deviation angle with a gradient threshold value to generate a pollution risk level; and S5, searching the multi-dimensional quality database matching characteristics based on the mutton variety attribution area, the muscle physical shear force value characterization quantity and the pollution risk level, and generating a comprehensive mutton quality classification result.
2. The method for classifying mutton quality based on multi-source data analysis according to claim 1, wherein the mutton variety attribution area comprises feature space decision boundary coordinates, clustering center geometric positions and variety confidence probability densities, the muscle physical shear force value characterization quantity comprises muscle fiber breaking strength indexes, connective tissue hardness parameters and meat tenderness quantification values, the element competition feature vector comprises ion channel occupation ratios, metabolism antagonism strength values and bioavailability competition indexes, the pollution risk level comprises heavy metal exposure toxicity levels, metabolism steady state damage gears and edible safety precaution categories, and the comprehensive mutton quality classification result comprises sensory flavor feature descriptions, nutritional value grading labels and commercial market grade codes.
3. The method for classifying mutton quality based on multi-source data analysis according to claim 2, wherein the step of obtaining the mutton variety home zone is specifically as follows: S111, acquiring mutton fatty acid detection data, calling a preset carbon chain length and double bond number orthogonal matrix, mapping the mutton fatty acid detection data into grid cells of the orthogonal matrix, performing concentration value aggregation summation on unsaturated isomers in each grid cell, and performing division operation on an aggregation summation value under a target chain length and saturated fatty acid concentration of a corresponding chain length to generate a chain specific unsaturation index; S112, based on the chain specific unsaturation index, arranging according to an incremental sequence of the number of carbon atoms, connecting discrete data points to generate a chain length-unsaturation distribution curve, performing first-order discrete difference operation on the chain length-unsaturation distribution curve, acquiring a change slope between adjacent nodes, extracting key nodes with the absolute value of the change slope being greater than a preset threshold, acquiring a slope value of the nodes and corresponding carbon chain position parameters, and generating a characteristic node topology mapping coordinate; S113, calling a support vector machine model, inputting the topological mapping coordinates of the characteristic nodes into the support vector machine model, mapping the coordinates to a high-dimensional characteristic space based on a preset kernel function, calculating the geometric interval between each sample point in the high-dimensional characteristic space and the optimal classification hyperplane, and carrying out region judgment according to the symbol attribute and the numerical range of the geometric interval to generate the mutton variety attribution region.
4. The method for classifying mutton quality based on multi-source data analysis according to claim 3, wherein the step of obtaining the muscle physical shear force value characterization quantity is specifically as follows: S211, collecting mutton near infrared spectrum data, scanning and positioning the central wavelength of a first-order frequency multiplication region characteristic peak corresponding to a water molecule O-H bond and the central wavelength of a frequency multiplication characteristic peak of a protein N-H bond in a full-band range, searching standard pure water spectrum peak reference wavelength and standard undegraded protein peak reference wavelength in a preset database, respectively executing subtraction difference operation of actual measurement characteristic peak wavelength and corresponding reference wavelength on water molecules and protein components, calculating the relative offset value of the two on a spectrum wavelength axis, and generating a spectrum characteristic peak frequency shift deviation set; S212, based on the spectrum characteristic peak frequency shift deviation set, extracting a water drift quantity value and a protein drift quantity value from the spectrum characteristic peak frequency shift deviation set as orthogonal coordinate components of a two-dimensional characteristic space, executing Euclidean distance operation in a vector space, solving the displacement modular length of a coordinate point relative to an origin, representing the dynamic separation degree of the spectrum response of water molecules relative to the spectrum response of protein, and constructing a water-protein decoupling coefficient; S213, calling a preset nonlinear piecewise mapping rule aiming at the water-protein decoupling coefficient, judging a tenderness conversion interval according to the numerical value of the water-protein decoupling coefficient, searching slope parameters and intercept parameters of the interval, and performing linear transformation and weighted summation operation 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, wherein the step of obtaining the element competition feature vector comprises the following steps: S311, acquiring a concentration set of essential trace elements of mutton and a concentration set of toxic heavy metals, calling a preset element atomic weight standard parameter library, retrieving relative atomic mass constants of corresponding elements of calcium, iron, zinc, lead, cadmium and arsenic, respectively executing division operation of mass concentration values and relative atomic mass constants on each essential trace element and toxic heavy metal, converting a concentration index taking mass as a reference into a mass concentration index of a substance taking particle number as a reference through unit conversion, and establishing a trace element and heavy metal molar concentration data set; S312, searching a preset biological antagonism competition channel mapping topology based on the microelement and heavy metal molar concentration data set, identifying and locking essential microelement and toxic heavy metal with competition ion channel occupation relation, pairing the locked elements pairwise to form antagonism element pairs, extracting essential microelement molar concentration of each antagonism element pair as a molecular item and toxic heavy metal molar concentration as a denominator item, performing point-to-point ratio calculation, quantifying the dose compression strength between the two pairs, and generating an antagonism element pair molar ratio set; s313, calling a preset feature vector dimension definition template aiming at the antagonistic element pair mole ratio set, sequentially mapping and filling each mole ratio value in the set into the corresponding dimension coordinate axis position of the multidimensional feature space according to the atomic number or chemical property liveness sequence of the element pair, and executing standardization processing based on the maximum and minimum values on the filled value sequence so as to eliminate the influence of the order-of-magnitude difference between the differential element pairs on the vector direction and construct the element competition feature vector.
6. The method for classifying mutton quality based on multi-source data analysis according to claim 5, wherein the step of obtaining the contamination risk level comprises the steps of: s411, calling a preset healthy mutton standard steady-state feature vector based on the element competition feature vector, executing vector point-to-point Euclidean distance operation in a multidimensional space, retrieving and extracting background ion interference intensity of a detection system, instrument zero reference offset and effective linear dynamic range of a feature space, and establishing a vector space geometric parameter set; S412, extracting corresponding dimension components of element competition feature vectors and healthy mutton standard steady-state feature vectors based on the vector space geometric parameter set, and calculating to obtain a metabolic steady-state deviation angle by combining background ion interference intensity, instrument zero reference offset, euclidean distance between vectors and effective linear dynamic range of feature space; s413, searching a preset pollution risk assessment grading standard aiming at the metabolic steady-state deviation angle, acquiring a gradient threshold sequence corresponding to a risk level, mapping the deviation angle value into a value interval defined by the gradient threshold sequence, judging the severity of heavy metal infection of a sample according to the interval position where the value falls, and generating a pollution risk level.
7. The method for classifying mutton quality based on multi-source data analysis according to claim 6, wherein the formula for obtaining metabolic steady-state deviation angle by calculation specifically comprises: ; Wherein, the Represents the metabolic steady-state departure angle, Representative element competition feature vector The normalized value of the dimension component, Standard steady state feature vector representing healthy mutton The normalized value of the dimension component, Representing the total number of dimensions of the feature vector, A normalized value representing background ion interference intensity of the detection system, A normalized value of the reference offset representing the instrument zero, Representing the normalized euclidean distance between the element competition feature vector and the healthy mutton standard steady state feature vector, Representing an effective linear dynamic range normalization value for the feature space.
8. The method for classifying mutton quality based on multi-source data analysis according to claim 7, wherein the step of obtaining the comprehensive mutton quality classification result specifically comprises the following steps: S511, calling a preset feature coding mapping protocol based on the mutton variety attribution area, the muscle physical shear force value characterization quantity and the pollution risk level, converting an area label into a space unique thermal coding vector, mapping the risk level into discrete ordinal weight factors, performing normalized scaling treatment on the shear force characterization quantity, and splicing and combining each item of processed data components according to a preset dimension sequence to establish a multi-source quality feature query vector; S512, searching a preset multi-dimensional quality database aiming at the multi-source quality feature query vector, traversing each type of standard quality template data stored in the database, respectively calculating the weighted Euclidean distance between the query vector and the standard template in a multi-dimensional feature space, extracting a candidate template set with the distance value smaller than a preset matching threshold, and converting the corresponding distance value into a similarity score through inverse proportion transformation to generate a quality feature matching degree matrix; And S513, based on the quality feature matching degree matrix, performing numerical sorting and maximum value retrieval of matrix array vectors, locking a feature template index with the highest similarity score, retrieving quality category definition data and attribute description fields associated with the indexes, mapping the data into specific text description and grade codes, and performing unified judgment on sample multi-dimensional attributes to generate a comprehensive mutton quality classification result.
9. A mutton quality classification system based on multi-source data analysis, wherein the system is for implementing a mutton quality classification method based on multi-source data analysis as claimed in any one of claims 1-8, the system comprising: The metabolism phenotype identification module is used for dividing an orthogonal matrix grid based on mutton fatty acid data, calculating concentration quotient, generating a chain-specific unsaturated degree index, arranging the chain-specific unsaturated degree index to construct a chain length-unsaturated degree distribution curve, extracting slope nodes of the chain length-unsaturated degree distribution curve, mapping the slope nodes to a high-dimensional characteristic space, and dividing the high-dimensional characteristic space by using a support vector machine to generate a mutton variety attribution area; The structure relaxation characterization module is used for collecting near infrared spectrum data of mutton, positioning characteristic peaks of water and protein, calculating a difference value between the characteristic peaks and standard wavelength, constructing the difference value into a two-dimensional vector, calculating the module length of the two-dimensional vector, constructing a water-protein decoupling coefficient, and generating a physical shear force value characterization quantity of muscle by utilizing a preset nonlinear piecewise mapping rule; The element antagonism calculation module is used for obtaining a concentration set of trace elements necessary for mutton and a concentration set of toxic heavy metals, calculating the mole ratio of antagonism elements to the concentration set, and polymerizing to construct an element competition characteristic vector; The pollution risk evaluation module is used for calling a healthy mutton standard steady-state feature vector based on the element competition feature vector, calculating cosine similarity with the element competition feature vector, solving a metabolic steady-state deviation angle, and comparing the metabolic steady-state deviation angle with a gradient threshold value to generate a pollution risk grade; And the quality grade mapping module is used for searching the matching characteristics of the multidimensional quality database based on the mutton variety attribution area, the muscle physical shear force value characterization quantity and the pollution risk grade to generate a comprehensive mutton quality classification result.

Description

Mutton quality classification method and system based on multi-source data analysis Technical Field 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. Background The technical field of quality classification relates to quantitative evaluation and grading treatment of multidimensional quality indexes such as physics, chemistry, biology and the like of objects such as agricultural products, foods, industrial products and the like, and the quantitative evaluation and grading treatment comprises key contents such as image analysis, spectrum detection, sensor data acquisition, feature extraction, quality evaluation standard construction, classification algorithm and the like. The development of the field depends on a precise sensing and evaluating mechanism for sample quality attributes, and generally combines a multidimensional detection means and an intelligent analysis technology to model sample attributes and grade quality, so that the method is widely applied to multiple application scenes such as food safety detection, agricultural product quality evaluation, industrial product quality control and the like, and has the characteristics of complex data sources, multiple evaluation dimensions and quantitative classification standards. The traditional mutton quality classification method refers to a mode of subjective judgment or single factor measurement on the characteristics of the mutton such as appearance, smell, texture, color, moisture content, fat distribution and the like through manual experience or a single detection means, and is usually judged by adopting a mode of manual classification, sensory evaluation or single data source based on near infrared spectrum, electronic nose, image recognition and the like. Such methods are generally operated based on specific types of data collected by a particular detection device, such as image data, for determining meat color and fat distribution, near infrared spectrum for detecting moisture and protein content, and electrochemical sensors for detecting odor components, and quality classification is performed with preset thresholds or simple rules. However, since a single data source has a limitation in the characterization capability, it is difficult to comprehensively reflect the multidimensional attribute of the mutton quality, and thus, there is a certain disadvantage in the classification accuracy and reliability. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a mutton quality classification method and a mutton quality classification system based on multi-source data analysis. In order to achieve the above purpose, the invention adopts the following technical scheme that the mutton quality classification method based on multi-source data analysis comprises the following steps: s1, dividing an orthogonal matrix grid based on mutton fatty acid data, calculating a concentration quotient, generating a chain-specific unsaturated degree index, arranging the chain-specific unsaturated degree index to construct a chain length-unsaturated degree distribution curve, extracting a slope node of the chain length-unsaturated degree distribution curve, mapping the slope node to a high-dimensional characteristic space, and dividing the high-dimensional characteristic space by using a support vector machine to generate a mutton variety attribution region; S2, collecting mutton near infrared spectrum data, positioning characteristic peaks of water and protein, calculating a difference value between the characteristic peaks and standard wavelength, constructing a two-dimensional vector by the difference value, calculating the module length of the two-dimensional vector, constructing a water-protein decoupling coefficient, and generating a physical shear force value characterization quantity of muscle by utilizing a preset nonlinear piecewise mapping rule; S3, acquiring concentration sets of trace elements and toxic heavy metal necessary for mutton, calculating the mole ratio of antagonistic elements to each other, and polymerizing to construct element competition characteristic vectors; s4, based on the element competition feature vector, calling a healthy mutton standard steady-state feature vector, calculating cosine similarity with the element competition feature vector, solving a metabolic steady-state deviation angle, and comparing the metabolic steady-state deviation angle with a gradient threshold value to generate a pollution risk level; and S5, searching the multi-dimensional quality database matching characteristics based on the mutton variety attribution area, the muscle physical shear force value characterization quantity and the pollution risk level, and generating a comprehensive mutton quality classification result. The improvement of the invention is that the mutton variety attribution