CN113125488B - Method for quickly identifying fat-filled artificial snowflake beef - Google Patents

Method for quickly identifying fat-filled artificial snowflake beef Download PDF

Info

Publication number
CN113125488B
CN113125488B CN202110428881.7A CN202110428881A CN113125488B CN 113125488 B CN113125488 B CN 113125488B CN 202110428881 A CN202110428881 A CN 202110428881A CN 113125488 B CN113125488 B CN 113125488B
Authority
CN
China
Prior art keywords
beef
fat
standard sample
peak
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110428881.7A
Other languages
Chinese (zh)
Other versions
CN113125488A (en
Inventor
李春保
李�浩
赵迪
周光宏
徐幸莲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Agricultural University
Original Assignee
Nanjing Agricultural University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing Agricultural University filed Critical Nanjing Agricultural University
Priority to CN202110428881.7A priority Critical patent/CN113125488B/en
Publication of CN113125488A publication Critical patent/CN113125488A/en
Application granted granted Critical
Publication of CN113125488B publication Critical patent/CN113125488B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N24/00—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects
    • G01N24/08—Investigating or analyzing materials by the use of nuclear magnetic resonance, electron paramagnetic resonance or other spin effects by using nuclear magnetic resonance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/30—Assessment of water resources

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • High Energy & Nuclear Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

The invention provides a method for quickly identifying fat-injected artificial snowflake beef by applying a low-field nuclear magnetic resonance technology, which adopts low-field nuclear magnetic resonance to detect natural and artificial snowflake beef with different fat injection quantities, and obtains T by using a specific CPMG sequence and inversion 2 And (3) distinguishing natural and artificial snowflake beef by using characteristic variables of a transverse relaxation time map and linear Support Vector Machine (SVM) and linear discrimination, further establishing a PLSR (partial least squares) and PCR (polymerase chain reaction) related prediction regression model of the fat filling amount aiming at the fat filling artificial snowflake beef, and verifying the reliability and accuracy of the model through a corresponding verification set. When the actual sample is measured, the fat injection amount of the fat injection artificial snowflake beef sample to be measured can be obtained only by substituting the measured characteristic variable value into the corresponding regression equation. The method of the invention lays a foundation for the rapid detection of the fat-filled artificial snowflake beef.

Description

Method for quickly identifying fat-filled artificial snowflake beef
Technical Field
The invention belongs to the field of food detection, and particularly relates to a method for rapidly identifying fat-filled artificial snowflake beef by applying a low-field nuclear magnetic resonance technology.
Background
With the rapid increase of economy, the income level of urban and rural residents is continuously improved, food culture and dietary structure are gradually improved, and the demand of people on high-quality and rich-nutrition meat products is continuously improved. In recent years, the beef consumption in China is on the whole in an increasing trend, and the beef consumption in China reaches 832.93 ten thousand tons. The beef with rich marbling (intramuscular fat), also called snowflake beef, has good taste and flavor, is popular with consumers, and has a price far higher than that of common beef. However, in actual production, the snowflake beef has very low yield and high production cost, which leads to the mass production of artificial snowflake beef, i.e. the beef fat or substitute (simulated fat emulsion) is evenly injected into the beef to be evenly distributed, and the processed beef is rich in the appearance characteristics and the taste of the snowflake beef by matching with certain tenderization and freezing wrapping processes. For example, chinese patent document CN 101961111A discloses a production method of snowflake beef, which comprises injecting emulsion prepared by using water, refined beef fat, starch, non-meat protein and the like as main raw materials into beef lean meat at 40-50 ℃ to form snowflake beef with alternate red and white colors, clear marbling and rich fat deposition; chinese patent document CN108925880A also discloses a snowflake beef production method. International patent WO 2020/143109 A1 discloses a snowflake beef production method, which comprises the steps of preparing base oil emulsion by using water, refined beef tallow, whey protein isolate, edible gelatin and the like as main raw materials, injecting the base oil emulsion into unfrozen beef by using a high-pressure injection machine at 55-60 ℃, and then carrying out processes of wrapping, quick-freezing and the like to prepare artificial snowflake beef with uniform and fine marbling, wherein the snowflake beef prepared by the method can present white fat patterns at low temperature (0-25 ℃), and is closer to natural snowflake beef in appearance structure. However, the artificial snowflake beef and the natural snowflake beef have differences in quality, nutrition, price and the like, so that great harm is brought to consumers.
A reliable method for identifying the artificial snowflake beef is imperative to be established. The detection research of snowflake beef mostly focuses on beef classification and fat content prediction through a computer vision technology and a near infrared technology, but the technologies cannot be applied to the identification of artificial snowflake beef, and the following two reasons mainly exist: firstly, computer vision and near infrared are analysis processing technologies based on beef section image information, excessive dependence on section visual information is avoided, and misjudgment probability is high due to the fact that real distribution conditions of marbling in a meat sample are ignored; secondly, with the continuous improvement and improvement of the fat injection formula and the technology, the artificial snowflake beef and the natural snowflake beef are almost difficult to be distinguished through appearance, and the computer vision and near infrared technology based on image processing and spectroscopy is difficult to be applied.
The low-field nuclear magnetic resonance is used as a novel rapid detection technology based on a sample component relaxation signal, and has wide application in the aspect of animal-derived food quality detection. The existing literature reports how to adopt low-field nuclear magnetic resonance technology to qualitatively or/and quantitatively determine water or fat in aquatic products such as yellow croakers, abalones, sea cucumbers, tilapia, caviar and the like (CN 109444199A, CN106018453A, CN105606637A, CN105548234A, liu Hongliang and the like, the biotechnology advances, 2020,10, 557-563), and the feasibility of the low-field nuclear magnetic resonance analysis of the whole individual is facilitated; for livestock and poultry products such as pork, beef and the like, the low-field nuclear magnetic resonance technology is widely used for water distribution and determination, but research reports for fat analysis are not found yet. No report on the detection of exogenous fat injection into meat is found. Theoretically, the low-field nuclear magnetic resonance technology can realize the rapid qualitative and quantitative judgment of the fat-infused artificial snowflake beef by analyzing fat molecule relaxation signals in a meat sample, and the difficulty is mainly how to reduce the influence of a sample moisture signal on fat analysis and how to distinguish signals of exogenous fat and inherent fat in the meat, thereby realizing the accurate quantitative identification of the fat-infused meat.
Disclosure of Invention
The invention is characterized in that the low-field nuclear magnetic resonance technology is applied to the quick qualitative and quantitative detection of the artificial snowflake beef for the first time, and the core is that the special metering analysis is carried out on the signals acquired by the low-field nuclear magnetic resonance analysis, so that the influence of the moisture signals of the sample on the fat analysis is reduced, and the distinguishing of the exogenous fat signals and the inherent component signals of the meat sample is realized; on the basis, a proper regression equation is constructed through regression model analysis and is used for predicting and calculating the content of the added exogenous fat. The method does not need to pretreat the sample, the testing process of the instrument is non-invasive, the use of chemical reagents is not involved, and the sample is not damaged.
The technical scheme of the invention is as follows:
a method for quickly identifying fat-injected snowflake beef comprises the following steps:
(1) Preparing a standard sample of the fat-injected artificial snowflake beef:
the standard samples comprise a natural beef standard sample and a fat-injected artificial snowflake beef standard sample; the standard sample of the fat-injected artificial snowflake beef is prepared by injecting fat into beef respectively according to different fat injection amounts; cutting the standard sample of the natural beef and the standard sample of the fat-filled artificial snowflake beef into a size suitable for being detected by a low-field nuclear magnetic resonance spectrometer;
(2) Extracting characteristic variables of the standard sample by low-field nuclear magnetic resonance:
respectively measuring a natural beef standard sample and fat-injected artificial snowflake beef standard samples with different fat injection amounts by using a low-field nuclear magnetic resonance spectrometer, and performing data inversion to obtain T of the natural beef standard sample and the fat-injected artificial snowflake beef standard samples with different fat injection amounts 2 A transverse relaxation time map; extracting information of each peak as a characteristic variable;
(3) Constructing a regression model: based on the information of each peak of the natural beef standard sample obtained in the step (2) and the fat-infused artificial snowflake beef standard samples with different fat infusion amounts as characteristic variables, establishing a Partial Least Squares Regression (PLSR) and/or a Principal Component Regression (PCR) model for obtaining the fat infusion amount aiming at the fat-infused artificial snowflake beef standard samples; the influence of the water signal of the sample on the fat analysis is reduced by combining corresponding characteristic parameters with chemometrics such as PCR, PLSR and the like.
(4) Detecting a sample to be detected: taking a beef sample to be detected, and extracting T of the beef sample to be detected under the same detection conditions in the step (2) 2 Distinguishing the signals of exogenous fat and inherent fat in meat by using the obtained characteristic variables of the transverse relaxation time spectrum through a linear Support Vector Machine (SVM) and a Linear Discriminant Analysis (LDA) method, and qualitatively analyzing whether the meat sample to be detected comprises the exogenous fat or not, wherein if the sample to be detected comprises the exogenous fat, the sample to be detected is the fat-injected snowflake beef; and (4) substituting the Partial Least Squares Regression (PLSR) and/or Principal Component Regression (PCR) model of the fat filling amount constructed in the step (3), and quantitatively identifying the fat filling amount of the beef sample to be detected.
Further, the beef used in the natural beef standard sample in the step (1) is the same as the sample to be detected in the step (4), preferably, the natural beef standard sample is beef; the beef used in the standard sample of the fat-injected artificial snowflake beef is the same as the sample to be detected in the step (4), preferably, the beef used in the standard sample of the fat-injected artificial snowflake beef is the beef back of the same kind as the sample to be detected in the step (4), and the injected fat is beef fat or a beef fat substitute (such as animal fat of pigs, sheep and the like, artificial fat and the like);
the range of the grease injection amount of the standard sample of the grease-injected artificial snowflake beef is 0-20% of the weight of the beef;
preferably, the fat injection amount of the fat injection artificial snowflake beef standard sample is that of beef quality: 0%, and one, two or more of 5%, 10%, 15%, 20%;
preferably, the syringe used for fat injection is a 5mL medical syringe, the specification of the injection needle is 1.6X 38.0mm, and beef tallow is melted at 70 ℃ during injection.
Further, the size suitable for the low-field nuclear magnetic resonance instrument in the step (1) is 3.0 × 2.0 × 1.5cm; preferably, the mass is 8.00 +/-0.01 g/block;
the cutting is in a direction perpendicular to the muscle fibers.
In a specific embodiment, the standard sample of natural beef is M9 and bovine eyeball meat and M12 and bovine eyeball meat, after fascia is removed, the beef is cut into meat slices with the thickness of about 2.0cm along the direction perpendicular to the muscle fiber, then the meat slices are taken and cut into strips with the thickness of about 3.0 × 1.5cm along the muscle fiber direction, so as to obtain meat blocks with the size of about 3.0 × 2.0 × 1.5cm, and the mass of the meat blocks is trimmed to 8.0g for standby.
The standard sample of the fat-injected artificial snowflake beef is prepared by taking fresh common beef back, removing fascia, cutting the back into meat slices with the thickness of about 2.0cm along the direction vertical to muscle fibers, then taking the meat slices, cutting the meat slices into strips with the thickness of about 3.0 multiplied by 1.5cm along the direction of the muscle fibers to obtain meat blocks with the size of about 3.0 multiplied by 2.0 multiplied by 1.5cm, trimming the mass of the meat blocks to 8.0g, melting refined beef tallow (namely beef fat) at 70 ℃, using a 5mL injector to match with a needle of a 1.6 multiplied by 38.0mm injector, sucking the melted beef tallow and injecting the melted beef tallow into the common beef back, wherein the fat injection amount is 0-20%, placing the fat-injected beef for 10min for waiting until the fat distribution is uniform, at the moment, cutting the beef from the position with the thickness of about 1.0cm in the middle, wherein the uniform marble patterns can be formed at the section, and the pattern density and the area of the standard sample of the fat-injected artificial snowflake beef are increased along with the increase of the fat injection amount.
Further, the step (2) comprises the following steps:
(2-1) setting of test parameters: calibrating a low-field nuclear magnetic resonance instrument, setting CPMG sequence parameters, and enabling a relaxation attenuation curve to be attenuated in a third area from an original point of an abscissa to a near-original point end of the abscissa, wherein the set inversion parameter range needs to contain all peaks of a sample, and adjacent peaks are required to be effectively distinguished;
(2-2) Standard sample test: respectively placing the natural beef standard sample and the fat-injected artificial snowflake beef standard sample in a nuclear magnetic tube, and collecting signals of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample at the center of a magnetic field to obtain relaxation attenuation curves of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample;
(2-3) inversion to obtain T 2 Transverse relaxation time spectra: the set inversion parameter range needs to contain all peaks of the standard sample, and adjacent peaks are effectively distinguished, and T of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample is obtained through inversion 2 A transverse relaxation time map;
(2-4) extraction of peak information: t from Natural beef Standard and fat-filled Artificial snowflake beef Standard 2 And extracting information of each peak from the transverse relaxation time map as a characteristic variable for subsequent analysis, wherein the information of each peak comprises peak starting time, peak top time, peak ending time, peak width, peak area and peak proportion.
Further, the low-field nuclear magnetic resonance apparatus in the step (2-1) is an NMI-20 type nuclear magnetic resonance imaging analyzer, the resonance frequency of the NMI-20 type nuclear magnetic resonance imaging analyzer is 21.3MHz, the magnet strength is 0.5 +/-0.08T, and the diameter of the probe coil is 40mm;
the instrument calibration was: before testing, ensuring that the radio frequency of an instrument is started, the temperature of a magnet is 32.00 ℃, placing the instrument into a standard oil sample used for correction, and after 30min of calibration, searching pulse width under an FID (Free indication Decay) sequence to determine the center frequency and drift frequency of the instrument, and the pulse width of 90 degrees (P1) and 180 degrees (P2), and completing instrument correction;
the CPMG sequence parameters are: the bandwidth SW =200kHz, the proton resonance frequency SF =21MHz, the RFD =0.002ms, the Tw is 2000-4000 ms, the analog gain RG1 is 15-20 db, the digital gain DRG1 is 0-5, the pre-amplification gain PRG is 0-3, the accumulation frequency NS is 8-32, the DR =1, the echo time TE is 0.15-0.25 ms, and the echo number NECH is 6000-10000;
preferably, the CPMG sequence parameters are: the receiver bandwidth SW =200kHz, the proton resonance frequency SF =21mhz, rfd =0.002ms, tw is 2500ms, the analog gain RG1 is 20db, the digital gain DRG1 is 2, the pre-amplification gain PRG is 1, the accumulation number NS is 16, dr =1, the echo time TE is 0.15ms, and the number of echoes NECH is 6000.
Further, in the step (2-2), the diameter of the nuclear magnetic tube is 30mm, the length of the nuclear magnetic tube is 200mm, and the acquisition of the signals is performed after the sample is placed in the nuclear magnetic tube and stabilized at the center of the magnetic field for 1 min.
Further, the inversion parameters in step (2-3) are: the number of sampling points is 200-400, the sampling range is 0.01-2000 ms, and the iteration number is 10 5 ～10 6 ，
Preferably, the inversion parameters are: the number of sampling points is 200, the sampling range is 0.01-2000 ms, and the iteration number is 10 6 。
Further, the T of the natural beef standard sample and the fat-filled artificial snowflake beef standard sample in the step (2-4) 2 The transverse relaxation time maps comprise four peaks respectively, 6 pieces of information of peak start time, peak top time, peak end time, peak width, peak area and peak proportion of each peak are extracted, and 24 characteristic variables in total are obtained.
Further, in the step (3), matlab R2020a software is used for analyzing to obtain the relationship between the principal component quantity, the dependent variable cumulative contribution rate, the independent variable interpretation degree and the root mean square error, and the principal component quantity is determined and extracted by utilizing the relationship between the principal component quantity and the dependent variable cumulative contribution rate and the relationship between the principal component quantity and the independent variable interpretation degree and the root mean square error; further obtaining a partial least squares regression model (PLSR) and/or a Principal Component Regression (PCR) model of the fat injection amount; the number of the principal components of the PCR model is 2 times of that of the PLSR model; preferably, the number of principal components of the PCR model is 6, and the number of principal components of the PLSR model is 3.
Further, the step (3) further comprises verifying the established regression model: and (3) establishing a verification set by adopting the same sample preparation method as the step (1) and the same test method as the step (2), extracting characteristic parameters, substituting the characteristic parameters into the established regression model to obtain a predicted value, obtaining a corresponding regression coefficient and a root mean square error through regression of the predicted value and the actual value of the verification set, and evaluating the accuracy of the established model, wherein when the regression coefficient reaches more than 0.8 and the root mean square error is less than 3, the regression model established in the step (3) can be used for rapidly identifying the fat-injected meat.
In a specific embodiment, the PLSR model in step (3) is: y1 (fat injection) = -45.02+ (-13.61 x T) 2b1s )+(-6.27*T 2b1t )+(-0.56*T 2b1e )+(0.29*W 2b1 )+(0.018*A 2b1 )+(-3.34*P 2b1 )+(-2.96*T 2b2s )+(5.56*T 2b2t )+(-0.26*T 2b2e )+(0.11*W 2b2 )+(0.024*A 2b2 )+(-9.19*P 2b2 )+(-0.46*T 21s )+(-0.16*T 21t )+(0.019*T 21e )+(0.045*W 21 )+(0.004*A 21 )+(-0.07*P 21 )+(0.0010*T 22s )+(0.0081*T 22t )+(0.0010*T 22e )+(0.0010*W 22 )+(0.0015*A 22 )+(-0.12*P 22 )；
Y2 (filling) = -65.45+ (-7.05 x T) 2b1s )+(-8.74*T 2b1t )+(-1.75*T 2b1e )+(-1.14*W 2b1 )+(0.018*A 2b1 )+(-3.018*P 2b1 )+(-3.21*T 2b2s )+(6.21*T 2b2t )+(0.064*T 2b2e )+(0.58*W 2b2 )+(0.024*A 2b2 )+(-8.61*P 2b2 )+(-0.45*T 21s )+(-0.10*T 21t )+(0.054*T 21e )+(0.071*W 21 )+(0.0039*A 21 )+(0.035*P 21 )+(0.0012*T 22s )+(0.0066*T 22t )+(0.00056*T 22e )+(4.16e-05*W 22 )+(0.00081*A 22 )+(-0.24*P 22 )。
The PLSR model and the PCR model can be directly applied to commercially available common beef (such as beef), and the detection and construction are carried out again from the step (1) for special beef (such as yak meat, yellow beef, angus beef and the like).
Further, the linear SVM and the linear discrimination in the step (4) are obtained by Classification Learner in Matlab R2020a software APP.
Compared with the existing snowflake beef detection method, the method has the following characteristics and advantages:
at present, except for the traditional physicochemical method, the detection of the snowflake beef is usually realized by an image and mechanical vision technology, but the existing method only aims at the classification of the snowflake beef and cannot realize the identification and the differentiation of the artificial fat-filled snowflake beef, and a scientific and effective method for identifying the artificial fat-filled snowflake beef is not reported temporarily.
(1) The quick identification method of the fat-filled artificial snowflake beef provided by the invention is based on the difference of low-field nuclear magnetic relaxation spectrums between the artificial snowflake beef and the natural snowflake beef, realizes the identification of the artificial snowflake beef and the natural snowflake beef by using a linear Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) method, and has high discriminant accuracy.
(2) In the method, the relaxation signals of the whole meat sample placed in the magnetic field can be measured by applying low-field nuclear magnetic resonance, the relaxation information of the whole components in the meat sample can be reflected, and the misjudgment probability is reduced.
(3) In the method, the use of chemical reagents is not involved in the testing and analyzing processes, the samples are not damaged, the testing time of a single sample is less than 1min, and the quick nondestructive analysis of the artificial fat-injected snowflake beef can be realized.
(4) In the method, for the extracted characteristic variables, a regression model is established by using algorithms such as PLSR (partial least squares) and PCR (polymerase chain reaction), the established model has high fitting degree and low root mean square error, and the established model can accurately predict the fat filling amount of the artificial fat filling snowflake beef through verification.
Drawings
FIG. 1 shows the marbling effect of the cut surface of artificial snowflake beef with different fat injection amount;
FIG. 2 is a graph of low-field NMR transverse relaxation time curves of natural snowflake beef (M9 and M12 and bull's eye beef) and artificial snowflake beef with different fat injection amounts;
FIG. 3 shows the results of discriminating natural snowflake beef (M9 and M12 and bull's eye beef) and fat-filled artificial snowflake beef based on linear SVM and Linear Discrimination (LDA);
FIG. 3A shows the linear SVM discrimination results of natural and fat-filled artificial snowflake beef;
FIG. 3B: linear discrimination results of natural and fat-injected artificial snowflake beef;
FIG. 3C: true rate (TPR) and False Negative Rate (FNR) graphs of SVM judgment results;
FIG. 3D: real rate (TPR) and False Negative Rate (FNR) graphs of LDA judgment results;
FIG. 3E: a positive prediction rate (PPV) and false occurrence rate (FDR) graph of SVM discrimination results;
FIG. 3F: positive prediction rate (PPV) and false occurrence rate (FDR) of LDA determination results.
FIG. 4 is a relationship between the number of extracted principal components and the cumulative contribution rate of dependent variables in the PLSR model;
FIG. 5 is a diagram showing the relationship between the number of extracted principal components in the PLSR model and the cumulative contribution of the independent variables and the root mean square error;
fig. 5a;
FIG. 5B: the root mean square error of the PLSR model is plotted as the quantity of extracted principal components.
FIG. 6 is a regression fit of the PLSR model validation set actual-predicted values;
FIG. 7 is a graph showing the relationship between the number of extracted principal components and the cumulative contribution of independent variables and the root mean square error in the PCR model;
FIG. 7A is a graph showing the cumulative contribution of an independent variable X of a PCR model as a function of the number of extracted principal components;
FIG. 7B: the PCR model root mean square error is plotted as the extracted principal component quantity.
FIG. 8 is a regression fit of the actual-predicted values of the validation set of the PCR model.
Detailed Description
The present method is described in detail below with reference to specific embodiments, but the present invention is not limited thereto.
Example 1:
1. purpose of experiment
The relaxation information of low-field nuclear magnetic resonance of the natural snowflake beef (and the bull eye beef) and the fat-filled artificial snowflake beef is obtained, and the natural snowflake beef (and the bull eye beef) and the fat-filled artificial snowflake beef are distinguished through a linear Support Vector Machine (SVM) method and a Linear Discriminant Analysis (LDA) method.
2. Laboratory apparatus
Low field nuclear magnetic resonance analyzer: NMI-20 model nuclear magnetic resonance imaging analyzer (resonance frequency 21.3MHz, magnet strength 0.5T, probe coil diameter 40 mm), shanghai Newman electronics technology Limited.
3. Experimental Material
Australia M9 and bovine eyeball meat, australia M12 and bovine eyeball meat, fresh common and bovine back meat, refined beef tallow;
4. experimental procedures
(1) Preparing a standard sample:
the standard sample of the natural beef is as follows: m9 and bull's eye flesh, M12 and bull's eye flesh, after removing fascia, cut into meat slices with the thickness of about 2.0cm along the direction perpendicular to muscle fibers, then take the meat slices and cut into strips with the thickness of about 3.0 multiplied by 1.5cm along the direction of the muscle fibers to obtain meat blocks with the size of about 3.0 multiplied by 2.0 multiplied by 1.5cm, then trim the mass to 8.0g, and place the cut meat samples on ice for standby.
The standard sample of the fat-injected artificial snowflake beef is as follows: taking fresh common beef back, removing fascia, cutting the fresh common beef back into meat slices with the thickness of about 2.0cm along the direction vertical to muscle fibers, then taking the meat slices, cutting the meat slices into strips with the thickness of about 3.0 multiplied by 1.5cm along the direction of the muscle fibers to obtain meat blocks with the size of about 3.0 multiplied by 2.0 multiplied by 1.5cm, trimming the meat blocks to 8.0g, melting refined beef tallow (namely beef fat) at 70 ℃, using a 5mL injector matched with a 1.6 multiplied by 38.0mm injector needle, sucking the molten beef tallow into the common beef back meat blocks, injecting the fat into the common beef back with the injection amount of 0% (namely common beef), 5%, 10%, 15% and 20%, placing the beef after fat injection for 10min, waiting for the fat to be uniformly distributed, at the moment, cutting the beef from the thick part of about 1.0cm in the middle, and showing uniform marble patterns at the section, and the density and the area of the standard sample of the snowflake fat injection artificial beef can be increased along with the increase of the fat injection amount. (as shown in fig. 1).
(2) Extracting characteristic variables of the standard sample by using field nuclear magnetic resonance:
(2-1) setting of test parameters:
the resonance frequency of an NMI-20 type nuclear magnetic resonance imaging analyzer is 21.3MHz, the strength of a magnet is 0.5 +/-0.08T, and the diameter of a probe coil is 40mm;
before testing, the radio frequency of the instrument is ensured to be started, the temperature of a magnet is 32.00 ℃, after the instrument is placed into a standard oil sample for sampling for 30min, the pulse width is searched under an FID sequence to determine the central frequency, the drift frequency and the pulse widths of 90 degrees (P1) and 180 degrees (P2), and the instrument correction is completed.
The parameters for switching the test sequence into the CPMG sequence are as follows: the receiver bandwidth SW =200kHz, the proton resonance frequency SF =21mhz, rfd =0.002ms, tw is 2500ms, the analog gain RG1 is 20db, the digital gain DRG1 is 2, the pre-amplification gain PRG is 1, the accumulation number NS is 16, dr =1, the echo time TE is 0.15ms, and the number of echoes NECH is 6000.
(2-2) Standard sample test:
after the instrument is corrected and stabilized, respectively placing M9 and M12 grade and natural beef standard samples of beef eye meat and standard samples of fat-injected artificial snowflake beef after the fat distribution is uniform in nuclear magnetic sample tubes with the diameter of 30mm and the length of 200mm, placing the nuclear magnetic tubes in the center of the magnetic field of the instrument, clicking 'cumulative sampling' to obtain relaxation attenuation curves of the samples (each sample is repeatedly measured for three times, and the average value is taken as the actual value corresponding to the sample).
(2-3) inversion to obtain T 2 Transverse relaxation time spectra:
after all samples are tested, all tested samples are selected in data query, batch inversion is selected, and inversion parameters are set as follows: the number of sampling points is 200, the sampling range is 0.01-2000 ms, and the iteration number is 10 6 And obtaining the T of the standard sample of the natural beef and the standard sample of the fat-injected artificial snowflake beef through inversion 2 Transverse relaxation time profiles (as shown in figure 2).
(2-4) extraction of peak information:
t obtained by inversion 2 The transverse relaxation time atlas discovers that four peaks basically appear in a natural beef standard sample and a fat-injected artificial snowflake beef standard sample, extracts information (peak starting time, peak vertex time, peak ending time, peak width, peak area and peak proportion) of each peak as characteristic variables, and comprises the following steps:
first peak-to-peak start time T 2b1s First peak-to-peak time T 2b1t First peak-to-peak end time T 2b1e First peak width W 2b1 First peak-to-peak area A 2b1 First peak-to-peak ratio P 2b1 ，
TABLE 1 information of the first relaxation peak of each beef sample
Figure GDA0003034441530000091
Second peak-to-peak start time T 2b2s Second peak-to-peak time T 2b2t Second peak-to-peak end time T 2b2e Second peak width W 2b2 Second peak-to-peak area A 2b2 Second peak-to-peak ratio P 2b2 ，
TABLE 2 second relaxation Peak information for each beef sample
Figure GDA0003034441530000092
Third peak to peak start time T 21s The third peak-to-peak time T 21t The third peak-to-peak end time T 21e Third peak width W 21 Third peak-to-peak area A 21 Third peak-to-peak ratio P 21 ，
TABLE 3 information of the third relaxation Peak for each beef sample
Figure GDA0003034441530000101
Fourth peak to peak start time T 22s The peak time T of the fourth peak 22t Fourth peak-to-peak end time T 22e Fourth peak width W 22 Fourth peak-to-peak area A 22 Fourth peak-to-peak ratio P 22 。
TABLE 4 information of the third relaxation Peak of each beef sample
Figure GDA0003034441530000102
(2-5) discrimination of meat type: t of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample with different proportions in the step (2-4) 2 Characteristic variables extracted from the transverse relaxation time spectrum are prediction variables, corresponding classifications are used as corresponding variables (the corresponding classifications refer to artificial snowflake beef, M9-grade and bulleye beef and M12-grade and bulleye beef), and a fat-injected artificial snowflake beef standard sample and a natural snowflake beef standard sample (the M9-grade and bulleye beef, the M12-grade and the bulleye beef) are distinguished by respectively using a linear SVM and Linear Discriminant (LDA) function in a Classification leaner in Matlab R2020a software APP. The results of the linear SVM and linear discriminant method (LDA) are shown in the attached figure 3 of the specification, which shows that the linear SVM and linear discriminant method (LDA) can qualitatively distinguish whether fat-infused snowflake beef is adopted (in the figure 3, fake is artificial snowflake beef, M9 represents M9 grade and bull's eye beef, and M12 is M12 grade and bull's eye beef).
(2-6) evaluation of differentiating Effect: and (3) evaluating the distinguishing results obtained by the two distinguishing methods (2-5) by adopting the real rate (TPR), the False Negative Rate (FNR), the positive prediction rate (PPV), the false occurrence rate (FDR) and the accuracy, wherein the distinguishing real rate of the two distinguishing methods for the fat-filled artificial snowflake beef reaches 100 percent, the distinguishing accuracy of each sample is higher than 98 percent, and the distinguishing of the standard sample of the natural beef (and the standard sample of the beef eyes) and the standard sample of the fat-filled artificial snowflake beef is realized.
Example 2:
1. purpose of experiment
And acquiring low-field nuclear magnetic resonance transverse relaxation time spectrums of the artificial snowflake beef with different grease injection quantities, establishing a PLSR (partial least squares) model for detecting the grease injection artificial snowflake beef by using low-field nuclear magnetic resonance and verifying the PLSR model.
2. Laboratory apparatus
The same as example 1;
3. experimental Material
Fresh beef back and refining beef tallow;
4. procedure of experiment
(1) Preparing artificial snowflake beef with different fat injection amounts: the same as example 1;
(2) The steps of extracting the characteristic variables of the standard sample by low-field nuclear magnetic resonance are the same as the example 1;
(3) Constructing a PLSR regression model and verifying the established PLSR regression model:
t obtained by inversion in step (2-3) 2 Finding that the fat-infused artificial snowflake beef basically has four peaks by a transverse relaxation time map, extracting information (peak starting time, peak vertex time, peak ending time, peak width, peak area and peak proportion) of each peak of the fat-infused artificial snowflake beef with different fat infusion amounts by the step (2-4) to be used as a characteristic variable T 2b1s 、T 2b1t 、T 2b1e 、W 2b1 、A 2b1 、P 2b1 、T 2b2s 、T 2b2t 、T 2b2e 、W 2b2 、A 2b2 、P 2b2 、T 21s 、T 21t 、T 21e 、W 21 、A 21 、P 21 、T 22s 、T 22t 、T 22e 、W 22 、A 22 、P 22 ，
(3-1) constructing a PLSR regression model based on the characteristic variables: analyzing the characteristic variables by using Matlab R2020a software to obtain the principal component quantity, the dependent variable accumulated contribution rate, the independent variable interpretation degree and the root mean square error, determining that the principal component quantity is 3 (as shown in figure 5) when the interpretation degree of the principal component quantity to the dependent variable is higher than 85% (figure 5A) and the root mean square error is not obviously changed along with the increase of the principal component quantity (figure 5B) by using the relation between the principal component quantity and the dependent variable accumulated contribution rate (as shown in figure 4) and the relation between the principal component quantity and the independent variable interpretation degree and the root mean square error, and further establishing an R PLSR model 2 Is 0.96, allThe square root error RMSE is 0.044, and then the corresponding regression equation is obtained as follows: y1 (fat injection) = -45.02+ (-13.61T 2b1s )+(-6.27*T 2b1t )+(-0.56*T 2b1e )+(0.29*W 2b1 )+(0.018*A 2b1 )+(-3.34*P 2b1 )+(-2.96*T 2b2s )+(5.56*T 2b2t )+(-0.26*T 2b2e )+(0.11*W 2b2 )+(0.024*A 2b2 )+(-9.19*P 2b2 )+(-0.46*T 21s )+(-0.16*T 21t )+(0.019*T 21e )+(0.045*W 21 )+(0.004*A 21 )+(-0.07*P 21 )+(0.0010*T 22s )+(0.0081*T 22t )+(0.0010*T 22e )+(0.0010*W 22 )+(0.0015*A 22 )+(-0.12*P 22 )。
(3-2) verification of the model:
the preparation method of the standard sample of the fat-injected artificial snowflake beef described in the step (1) in the embodiment 1 is adopted to inject fat into the fresh beef back to obtain a fat-injected beef sample, a verification set is established according to the test methods of the steps (2-1) to (2-4) in the embodiment 1, 24 pieces of information of 4 peaks are extracted as characteristic variables to be substituted into the regression equation of the PLSR model to obtain a predicted value, the regression of the predicted value and the actual value of the verification set is used for obtaining a corresponding regression coefficient and a root mean square error, and the regression fitting degree R of the predicted value-the actual value is obtained by evaluating the accuracy of the established model (as shown in figure 6) 2 0.95, root mean square error 9.28 (table 5).
Table 5 verification set PLSR model prediction value-actual value comparison table of artificial snowflake beef samples
Figure GDA0003034441530000121
Figure GDA0003034441530000131
(4) Prediction of meat samples to be tested
Cutting a certain mass of meat sample to be tested into meat blocks with the size of about 3.0 multiplied by 2.0 multiplied by 1.5cm, trimming the mass of the meat blocks to 8.0g, putting the meat blocks into a low-field nuclear magnetic resonance spectrometer, testing and inverting the meat blocks under the same conditions of the steps (2-1) to (2-4), extracting characteristic parameters of a transverse relaxation time spectrum of the meat blocks, and qualitatively analyzing the meat sample to be tested as fat-filled snowflake beef by using a linear SVM and a linear discrimination method according to obtained characteristic variables; the characteristic variables are substituted into the regression equation of the PLSR model obtained in the embodiment (3-1), and the fat injection amount of the PLSR model can be predicted through the model (the result is shown in Table 6), so that qualitative and quantitative identification of the fat injection artificial snowflake beef is realized.
TABLE 6 comparison table of PLSR model predicted value and actual value of artificial snowflake beef sample
Figure GDA0003034441530000132
Example 3:
1. purpose of experiment
And acquiring low-field nuclear magnetic resonance transverse relaxation time spectrums of the artificial snowflake beef with different fat injection quantities, establishing a PCR (polymerase chain reaction) model for detecting the fat injection artificial snowflake beef by low-field nuclear magnetic resonance, verifying the PCR model, and comparing the PCR model with a PLSR (partial least squares regression) model.
2. Laboratory apparatus
The same as example 1;
3. experimental Material
The same as example 2;
4. procedure of experiment
(1) Preparing artificial snowflake beef with different fat injection amounts: as in example 1;
(2) The steps of extracting the characteristic variables of the standard sample by low-field nuclear magnetic resonance are the same as those in example 1;
(3) Constructing a PCR regression model and verifying the established PCR regression model:
similarly, for the fat-filled artificial snowflake beef T with different fat-filled amounts 2 Information of four peaks (peak start time, peak top time, peak end time, peak width, peak area and peak proportion) of transverse relaxation time map is extracted as characteristic variable (T) 2b1s 、T 2b1t 、T 2b1e 、W 2b1 、A 2b1 、P 2b1 、T 2b2s 、T 2b2t 、T 2b2e 、W 2b2 、A 2b2 、P 2b2 、T 21s 、T 21t 、T 21e 、W 21 、A 21 、P 21 、T 22s 、T 22t 、T 22e 、W 22 、A 22 、P 22 )，
(3-1) constructing a PCR regression model based on the characteristic variables: analyzing the characteristic variables by using Matlab R2020a software to obtain the number of principal components, the accumulated contribution rate of dependent variables, the interpretation degree of independent variables and the root mean square error, and determining the number of extracted principal components by using the relationship between the number of principal components and the relationship between the dependent variables and the root mean square error, namely determining the number of extracted principal components to be 6 (as shown in FIG. 7) when the interpretation degree of the number of principal components to the dependent variables is higher than 85% (FIG. 7A) and the root mean square error is not obviously changed any more with the increase of the number of principal components (FIG. 7B), and further establishing a PCR model, wherein R of the established model is R 2 Is 0.95, the root mean square error RMSE is 0.061, and the corresponding regression equation is obtained as:
y2 (filling) = -65.45+ (-7.05 x T) 2b1s )+(-8.74*T 2b1t )+(-1.75*T 2b1e )+(-1.14*W 2b1 )+(0.018*A 2b1 )+(-3.018*P 2b1 )+(-3.21*T 2b2s )+(6.21*T 2b2t )+(0.064*T 2b2e )+(0.58*W 2b2 )+(0.024*A 2b2 )+(-8.61*P 2b2 )+(-0.45*T 21s )+(-0.10*T 21t )+(0.054*T 21e )+(0.071*W 21 )+(0.0039*A 21 )+(0.035*P 21 )+(0.0012*T 22s )+(0.0066*T 22t )+(0.00056*T 22e )+(4.16e-05*W 22 )+(0.00081*A 22 )+(-0.24*P 22 )。
Compared with the PLSR model, the established PCR model achieves the fitting degree and the root mean square error which are equivalent to those of the PLSR model under the condition of extracting more principal components by 1 time.
(3-2) verification of the model:
fat injection is carried out on fresh beef back ridge meat by adopting the same sample preparation method as the preparation method of the fat injection artificial snowflake beef standard sample in the step (1) of the example 1 to obtain fat injectionA fat beef sample is established, a verification set is established according to the testing method in the step (2) in the embodiment 1, 24 characteristic variables of 4 peaks are extracted and are substituted into the regression equation of the PCR model to obtain a predicted value, corresponding regression coefficients and root mean square errors are obtained by regression of the predicted value and the actual value of the verification set, the accuracy of the established model is evaluated (as shown in figure 8), and the regression fitting degree R of the obtained predicted value-actual value 2 It was 0.90, with a root mean square error of 9.83 (Table 7).
Table 7 verification set PCR model predicted value-actual value comparison table of artificial snowflake beef samples
Figure GDA0003034441530000141
/>
Figure GDA0003034441530000151
(4) Prediction of meat samples to be tested
Cutting a certain mass of meat sample to be tested into meat blocks with the size of about 3.0 multiplied by 2.0 multiplied by 1.5cm, trimming the mass of the meat blocks to 8.0g, putting the meat blocks into a low-field nuclear magnetic resonance spectrometer, testing and inverting the meat blocks under the same conditions of the steps (2-1) to (2-4), extracting characteristic parameters of a transverse relaxation time spectrum of the meat blocks, and qualitatively analyzing the meat sample to be tested as fat-filled snowflake beef by using a linear SVM and a linear discrimination method according to obtained characteristic variables; the characteristic variables are substituted into the regression equation of the PCR model obtained in the embodiment (3-1), and the fat injection amount of the PCR model can be predicted through the model (the result is shown in Table 8), so that the qualitative and quantitative identification of the fat injection artificial snowflake beef is realized.
Table 8 comparison table of predicted values and actual values of PCR model of artificial snowflake beef samples
Figure GDA0003034441530000152
In the shown case, under given detection and inversion parameters, by extracting information of each peak in a low-field nuclear magnetic resonance transverse relaxation time spectrum of a corresponding sample, firstly, the differentiation of fat-injected artificial snowflake beef and natural snowflake beef (and bull-eye beef) is realized by a linear Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) method, namely, the differentiation of exogenous fat signals and meat sample inherent component signals is realized; furthermore, the prediction analysis of the fat filling amount of the fat filling artificial snowflake beef is effectively realized by combining the PLSR and the PCR model, in the embodiment shown in the method, the linear SVM method realizes 100% of distinction of the fat filling artificial snowflake beef compared with a Linear Discriminant (LDA) method, and the linear SVM method (99.1%) is slightly higher than the Linear Discriminant (LDA) method (98.4%) in the integral discriminant accuracy. In the aspect of predicting the fat filling amount of the artificial fat filling snowflake beef, compared with a PCR model, the PLSR model has higher fitting degree and lower root mean square error in extracting less main components than the PLSR model, and the result of a verification set shows that the fitting degree of regression of a PLSR model predicted value and an actual value reaches 0.95 and is far higher than 0.90 of the PCR model, which shows that the accuracy of the PLSR model for the predicted result of the fat filling amount of the artificial fat filling snowflake beef is obviously better than that of the PCR, the reliability of the model is higher, and the PLSR model is more suitable for quickly, qualitatively and quantitatively analyzing the fat filling artificial snowflake beef.
The above examples are given for the purpose of illustration only, and are not intended to limit the scope of the claims, and all equivalent variations which characterize the methods of the present patent are within the scope of the claims.
The method for preparing the fat-injected artificial snowflake beef sample comprises the steps of taking pure refined beef tallow as base material oil for injection, and modeling and predicting by adopting the method after the formula of the injected base material oil is adjusted on the basis of the refined beef tallow, so that the method still falls within the protection scope of the invention.

Claims (10)

1. A method for quickly identifying fat-injected snowflake beef is characterized by comprising the following steps:
(1) Preparing a standard sample:
the standard samples comprise a natural beef standard sample and a fat-injected artificial snowflake beef standard sample; the standard sample of the fat-injected artificial snowflake beef is prepared by injecting fat into beef respectively with different fat injection amounts; cutting the standard sample of the natural beef and the standard sample of the fat-filled artificial snowflake beef into a size suitable for being detected by a low-field nuclear magnetic resonance spectrometer;
(2) Extracting characteristic variables of the standard sample by low-field nuclear magnetic resonance:
respectively measuring a natural beef standard sample and fat-injected artificial snowflake beef standard samples with different fat injection amounts by using a low-field nuclear magnetic resonance spectrometer, and performing data inversion to obtain T of the natural beef standard sample and the fat-injected artificial snowflake beef standard samples with different fat injection amounts 2 A transverse relaxation time map; extracting information of each peak as a characteristic variable;
(3) Constructing a regression model: based on the information of each peak of the natural beef standard sample obtained in the step (2) and the fat-injected artificial snowflake beef standard samples with different fat injection amounts as characteristic variables, establishing a partial least square regression model and/or a principal component regression model of the fat injection amount aiming at the fat-injected artificial snowflake beef standard samples;
(4) Detecting a sample to be detected: taking a beef sample to be detected, and extracting T of the beef sample to be detected under the same detection conditions in the step (2) 2 Carrying out qualitative analysis on the characteristic variables of the transverse relaxation time map to determine whether the meat sample to be tested is fat-filled snowflake beef or not by applying a linear Support Vector Machine (SVM) and a linear discrimination method to the obtained characteristic variables; if the meat sample to be detected is the fat-injected snowflake beef, substituting the characteristic variables into the partial least square regression and/or principal component regression model of the fat injection amount constructed in the step (3), and quantitatively identifying to obtain the fat injection amount of the meat sample of the beef to be detected;
the beef used in the natural beef standard sample in the step (1) is the same as the sample to be detected in the step (4);
the step (2) comprises the following steps:
(2-1) setting of test parameters: calibrating a low-field nuclear magnetic resonance instrument, setting CPMG sequence parameters, and enabling a relaxation attenuation curve to be attenuated in a third area from an original point of an abscissa to a near-original point end of the abscissa, wherein the set inversion parameter range needs to contain all peaks of a sample, and adjacent peaks are required to be effectively distinguished;
(2-2) Standard sample test: respectively placing the natural beef standard sample and the fat-injected artificial snowflake beef standard sample in a nuclear magnetic tube, and collecting signals of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample at the center of a magnetic field to obtain relaxation attenuation curves of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample;
(2-3) inversion to obtain T 2 Transverse relaxation time spectra: the set inversion parameter range needs to contain all peaks of the standard sample, and adjacent peaks are effectively distinguished, and T of the natural beef standard sample and the fat-injected artificial snowflake beef standard sample is obtained through inversion 2 A transverse relaxation time map;
(2-4) extraction of peak information: t from Natural beef Standard sample and fat-filled Artificial snowflake beef Standard sample 2 Extracting information of each peak from the transverse relaxation time map as a characteristic variable of subsequent analysis, wherein the information of each peak comprises peak starting time, peak top time, peak ending time, peak width, peak area and peak proportion;
the low-field nuclear magnetic resonance analyzer in the step (2-1) is an NMI-20 type nuclear magnetic resonance imaging analyzer, the resonance frequency of the NMI-20 type nuclear magnetic resonance imaging analyzer is 21.3MHz, the strength of a magnet is 0.5 +/-0.08T, and the diameter of a probe coil is 40mm;
the partial least squares regression model in the step (3) is as follows: y1 (fat injection) = -45.02+ (-13.61T 2b1s )+(-6.27*T 2b1t )+(-0.56*T 2b1e )+(0.29*W 2b1 )+(0.018*A 2b1 )+(-3.34*P 2b1 )+(-2.96*T 2b2s )+(5.56*T 2b2t )+(-0.26*T 2b2e )+(0.11*W 2b2 )+(0.024*A 2b2 )+(-9.19*P 2b2 )+(-0.46*T 21s )+(-0.16*T 21t )+(0.019*T 21e )+(0.045*W 21 )+(0.004*A 21 )+(-0.07*P 21 )+(0.0010*T 22s )+(0.0081*T 22t )+(0.0010*T 22e )+(0.0010*W 22 )+(0.0015*A 22 )+(-0.12*P 22 )；
The principal component regression model in the step (3) is as follows: y2 (filling) = -65.45+ (-7.05 x T) 2b1s )+(-8.74*T 2b1t )+(-1.75*T 2b1e )+(-1.14*W 2b1 )+(0.018*A 2b1 )+(-3.018*P 2b1 )+(-3.21*T 2b2s )+(6.21*T 2b2t )+(0.064*T 2b2e )+(0.58*W 2b2 )+(0.024*A 2b2 )+(-8.61*P 2b2 )+(-0.45*T 21s )+(-0.10*T 21t )+(0.054*T 21e )+(0.071*W 21 )+(0.0039*A 21 )+(0.035*P 21 )+(0.0012*T 22s )+(0.0066*T 22t )+(0.00056*T 22e )+(4.16e-05*W 22 )+(0.00081*A 22 )+(-0.24*P 22 )。
2. The method for rapidly identifying the fat-injected snowflake beef as claimed in claim 1, wherein the beef used in the natural beef standard sample is beef and bull's eye beef; the beef used in the standard sample of the fat-injected artificial snowflake beef is the same as the sample to be detected in the step (4), and the injected fat is beef fat or a beef fat substitute;
the range of the grease injection amount of the standard sample of the grease-injected artificial snowflake beef is 0-20% of the mass of the beef.
3. The method for rapidly identifying fat-filled snowflake beef as claimed in claim 1, wherein the suitable size detected by the low-field nuclear magnetic resonance apparatus in step (1) is 3.0 x 2.0 x 1.5cm.
4. The method for rapidly identifying fat filled snowflake beef of claim 1, wherein the instrumental calibration is: before testing, ensuring that the radio frequency of the instrument is started, the temperature of a magnet is 32.00 ℃, and after calibration is carried out for 30min, searching pulse width under an FID sequence to determine the central frequency, drift frequency and pulse width of 90 degrees and 180 degrees, and completing instrument correction;
the CPMG sequence parameters are: the bandwidth SW =200kHz, the proton resonance frequency SF =21MHz, the RFD =0.002ms, the Tw is 2000-4000 ms, the analog gain RG1 is 15-20 db, the digital gain DRG1 is 0-5, the pre-amplification gain PRG is 0-3, the accumulation frequency NS is 8-32, the DR =1, the echo time TE is 0.15-0.25 ms, and the echo number NECH is 6000-10000.
5. The method for rapidly identifying the fat-injected snowflake beef as claimed in claim 1, wherein in the step (2-2), the diameter of the nuclear magnetic tube is 30mm, the length of the nuclear magnetic tube is 200mm, and the collection of the signals is carried out after the sample is placed in the nuclear magnetic tube and is stabilized at the center of a magnetic field for 1 min.
6. The method for rapidly identifying the fat filled snowflake beef as claimed in claim 1, wherein the inversion parameters in the step (2-3) are as follows: the number of sampling points is 200-400, the sampling range is 0.01-2000 ms, and the iteration number is 10 5 ～10 6 。
7. The method for rapidly identifying fat-filled snowflake beef as claimed in claim 1, wherein the T of the natural beef standard sample and the fat-filled artificial snowflake beef standard sample in the step (2-4) 2 The transverse relaxation time maps comprise four peaks respectively, and the peak start time, the peak top time, the peak end time, the peak width, the peak area and the peak proportion of each peak are extracted to obtain 24 characteristic variables in total.
8. The method for rapidly identifying the fat-injected snowflake beef according to claim 1, wherein the step (3) is to use Matlab R2020a software to perform analysis to obtain the principal component quantity, the dependent variable cumulative contribution rate, the independent variable explainability and the root mean square error, and determine the extracted principal component quantity by using the relationship between the principal component quantity and the dependent variable cumulative contribution rate and the relationship between the principal component quantity and the independent variable explainability and the root mean square error; further obtaining a partial least squares regression model and/or a principal component regression model of the grease injection amount; the principal component quantity of the principal component regression model is 2 times of the principal component quantity of the partial least square regression model.
9. The method for rapidly identifying the fat-filled snowflake beef as claimed in claim 1, wherein the step (3) further comprises the step of verifying the established regression model: and (3) establishing a verification set by adopting the sample preparation method same as the step (1) and the testing method same as the step (2), extracting characteristic parameters, substituting the characteristic parameters into the established regression model to obtain a predicted value, obtaining a corresponding regression coefficient and a root mean square error through regression of the predicted value and the actual value of the verification set, and evaluating the accuracy of the established model, wherein when the regression coefficient reaches more than 0.8 and the root mean square error is less than 3, the regression model established in the step (3) can be used for rapidly identifying the fat-injected meat.
10. The method for rapidly identifying fat-infused snowflake beef as claimed in claim 1, wherein the linear SVM and linear discrimination in step (4) are obtained by Classification Learner in Matlab R2020a software APP.
CN202110428881.7A 2021-04-21 2021-04-21 Method for quickly identifying fat-filled artificial snowflake beef Active CN113125488B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110428881.7A CN113125488B (en) 2021-04-21 2021-04-21 Method for quickly identifying fat-filled artificial snowflake beef

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110428881.7A CN113125488B (en) 2021-04-21 2021-04-21 Method for quickly identifying fat-filled artificial snowflake beef

Publications (2)

Publication Number Publication Date
CN113125488A CN113125488A (en) 2021-07-16
CN113125488B true CN113125488B (en) 2023-04-18

Family

ID=76778439

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110428881.7A Active CN113125488B (en) 2021-04-21 2021-04-21 Method for quickly identifying fat-filled artificial snowflake beef

Country Status (1)

Country Link
CN (1) CN113125488B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021037913A1 (en) * 2019-08-27 2021-03-04 Nanonord A/S A method of and a system for determining fat concentration in a flowable sample by nuclear magnetic resonance

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
NO311111B1 (en) * 1998-04-03 2001-10-08 Geir H Soerland Method for determining the amount of fat and water in a biological sample
JP5294230B2 (en) * 2007-02-23 2013-09-18 独立行政法人産業技術総合研究所 Identification and quantification method of marbling and non-invasive measuring apparatus of marbling
CN105510371A (en) * 2015-11-30 2016-04-20 上海纽迈电子科技有限公司 Low-field nuclear magnetic resonance-based detection method of fat content for living animals
CN105548234A (en) * 2015-12-08 2016-05-04 大连工业大学 Method for nondestructive detection of water and fat contents of yellow croaker
CN109444199A (en) * 2018-10-22 2019-03-08 大连工业大学 Utilize the refrigeration degree of beef freshness lossless detection method of low-field nuclear magnetic resonance technology

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2021037913A1 (en) * 2019-08-27 2021-03-04 Nanonord A/S A method of and a system for determining fat concentration in a flowable sample by nuclear magnetic resonance

Also Published As

Publication number Publication date
CN113125488A (en) 2021-07-16

Similar Documents

Publication Publication Date Title
Wang et al. Quality analysis, classification, and authentication of liquid foods by near-infrared spectroscopy: A review of recent research developments
Cheng et al. Marbling analysis for evaluating meat quality: Methods and techniques
CN109406740A (en) A kind of evaluation method of freshwater fish quality
Choi et al. Correlations of sensory quality characteristics with intramuscular fat content and bundle characteristics in bovine longissimus thoracis muscle
CN105758884A (en) Method for identifying sea cucumber drying modes and method for monitoring sea cucumber quality change during drying
CN109444199A (en) Utilize the refrigeration degree of beef freshness lossless detection method of low-field nuclear magnetic resonance technology
Lee et al. Determination of intramuscular fat content in beef using magnetic resonance imaging
CN113125488B (en) Method for quickly identifying fat-filled artificial snowflake beef
Albrecht et al. A new technique for objective evaluation of marbling in beef
CN106556680A (en) It is a kind of based on Electronic Nose Technology fast prediction meat and its method for product fat oxidation degree
CN106841265B (en) A kind of pork kills postcooling Moisture Movement detection method
Gai et al. Effects of Hydrocolloid Injection on the Eating Quality of Pork Analyzed Based on Low‐Field Nuclear Magnetic Resonance (LF‐NMR)
Wold et al. NIR spectroscopic techniques for quality and process control in the meat industry
AU2015259369B2 (en) Method and system of assessing or analyzing muscle characteristics including strength and tenderness using ultrasound
CN111122492B (en) Near-infrared detection-based fast water-injected meat screening method
CN111990331B (en) Living body breeding method for pork streaky pork
Ilic et al. How do we eat meat–the role of structure, mechanics, oral processing, and sensory perception in designing meat analogs
Song et al. Objective meat quality from quality grade and backfat thickness of Hanwoo steers
CN111915564A (en) Frozen meat product optical characteristic detection method combining machine vision and OCT technology
Song et al. A novel non-destructive manner for quantitative determination of plumpness of live Eriocheir sinensis using low-field nuclear magnetic resonance
CN109613190A (en) A kind of meat product type and quartile length method based on amino acid composition
CN109409687A (en) Method based on apple feedstock specifications prediction dried product Color Quality
Tang et al. Determination of fat content in pork by low-field nuclear magnetic resonance.
CN114354670A (en) Water and glue injection meat detection method based on MLP neural network
CN108344779B (en) Method for rapidly detecting cow recessive mastitis grade based on dielectric spectrum technology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant